CN101889442B - System for combining a plurality of views of real-time streaming interactive video - Google Patents

Abstract

An apparatus comprises a plurality of servers arranged in a parallel processing configuration to create a plurality of streams of low-latency streaming interactive video. Each server produces a stream of low-latency streaming interactive video. A unit combines the plurality of streams into a new stream of low-latency streaming interactive video. The unit is also operable to provide a user interface that allows real-time interactive manipulation of the resulting combined streams by a user.

Description

For the system of the view in conjunction with multiple real-time streaming interactive videos

Related application

The application be that on December 10th, 2002 submits to, be entitled as " for the device and method (Apparatus and Method for Wireless Video Gaming) of wireless video game ", the 10/315th, the part of No. 460 continues (CIP) application, and this application has transferred the assignee of this CIP application.

Technical field

The disclosure relates generally to the field of the data handling system of the ability of improving user's operation and access audio and video medium.

Background technology

Since Thomas Edison (Thomas Edison) is since the epoch, the audio frequency of record and film media have become social one side.Early in the twentieth century, the audio frequency media of record (magnetic post and disc) and film media (jukebox and film) are extensively issued, but two kinds of technology are still in its starting stage.In the later stage twenties 20th century, on the basis of general marketplace, by film and audio frequency combination, afterwards color film and audio frequency are combined.Radio broadcasting develops into the broadcast general marketplace audio frequency media of the form of supporting to a great extent advertisement gradually.In the time setting up TV (TV) broadcast standard in the mid-40 in 20th century, TV engages with the form of broadcast general marketplace media with radio, thereby film precedence record or live is brought in family.

Till mid-term in 20th century, most of American family had the disc player (phonograph record player) of the audio frequency media for playing record, for receiving the wireless device of live broadcast audio, and for playing the television set of live broadcast audio/video (A/V) media.Usually described 3 " media players " (disc player, wireless device and TV) is combined in the cabinet of total public loud speaker, becomes " media center " of family.Although media selection is limited for consumers, media " ecosystem " are highly stable.Most consumers is known the four corner that how to use " media player " and can enjoy its ability.Simultaneously, the publisher of media (is mostly film and television operations chamber, and music company) can be by its media distribution to cinema and family, and do not suffer pirate or " two time selling " (yet, having used the sale again of media) widely.Conventionally, publisher can not taken in from two time selling, and therefore, two time selling has reduced the income that publisher obtains from the buyer of the use media of originally can controlling oneself for new sale.Although 20th century interim period really have the disc that used and sell, but described sale can not have large impact to disc publisher, because be different from film or video frequency program (its conventionally watched once by adult or only for several times), music track can be listened to hundreds of times or even thousands of times.Therefore, music media is more than movie/video media " prolonged " (also,, for the consumer that grows up, it has lasting value).Once buy disc, if consumer likes this music, consumer may be kept for a long time.

From mid-term in 20th century till now, the media ecosystem has all experienced a series of radical changes for the interests of consumer and publisher and loss.In the case of the extensive introducing of audio sound-recording machine (especially thering is the stereosonic cassette tape of high-quality), the consumer's facility that truly has higher degree.But it also indicates now consumer's media practice widely-pirate beginning.Really, many consumers record its own disc for using box type magnetic to bring for the purpose of facility purely, for example, copy but the consumer who day by day increases (, preparing the student of access disc collection each other in dormitory) will carry out piracy.And consumer is the music of recording via radio play, but not buy disc or tape from publisher.

The appearance of consumer VCR causes more consumer's facility, because VCR is set to record TV program now, it can be watched in time after a while, and VCR also causes the foundation of video rental industry, and wherein film and the design of TV program can be at " program request " basic enterprising line access.The fast Development of the general marketplace home media equipment since 20th century the mid-80s has caused consumer's unprecedented selection and convenience degree, and also causes media to publish the Rapid Expansion in market.

Now, consumer faced multimedia and selected and crossed multimedia equipment, wherein many media or specific publishers that are bound to particular form.The media consumer's of making earnest efforts may, by a pile equipment connection to TV and computer in each room, house, cause " mouse nest formula " cable and a group Long-distance Control to one or more television set and/or personal computer (PC).(in the application's context, term " personal computer " or " PC " refer to the computer that is suitable for any kind using in family or office), comprises desktop computer, Macintosh (Macintosh machine) or other non-Windows (form) computer, the equipment compatible with Windows, Unix variant, notebook etc.).Described equipment can comprise that video game console, VCR, DVD player, Audio Loop are around sound effect processor/amplifier, satellite set top box, cable tv Set Top Box etc.In addition,, for the consumer who makes earnest efforts, may there is due to consistency problem the equipment of multiple similar functions.For example, consumer may have HD-DVD and blue light (Blu-ray) DVD player, or Microsoft Xbox (Microsoft's home game machine) with Sony Playstation (game station of Sony) video game system both.In fact, due to some across game console version the incompatibility of game, consumer may have XBox and version after a while (such as, Xbox ) both.Frequently, consumer is for using which video inputs and which far-end to be puzzled.Even for example, by disk holding in correct player (, DVD, HD-DVD, blue light, Xbox or Playstation) in, select for video and the audio input end of this equipment and after finding correct Long-distance Control, consumer still faces technological challenge.For example, under the situation of widescreen DVD, user may need first to determine correct aspect ratio (for example, 4: 3, completely, amplification, wide amplification, cinema be wide etc.) and then on its TV or monitor screen, set correct aspect ratio.Similarly, user for example may need first to determine correct Audio Loop, around sound effect system form (, AC-3, Dolby Digital, DTS etc.) and then set correct Audio Loop around sound effect system form.Often, consumer do not recognize media content under its whole abilities of may not enjoying its TV or audio system (for example, watch the film with wrong aspect ratio extruding, or listen to stereosonic audio frequency but not around the audio frequency of audio).

Day by day increase ground, will add in the storehouse of equipment with the media device based on the Internet.Similar Sonos (George Soros) the audio frequency apparatus of digital music system makes directly from the Internet mobile (stream) of audio frequency.Similarly, similar Slingbox ^tM(depending on Lingbao City ^tM) amusement player equipment records video and make it flow via home network or flow and go out via the Internet, wherein can be on PC this video of remote watching.And IPTV (IPTV) service via Digital Subscriber Line (DSL) or other home interconnect nets connect and the service of analogous cable TV are provided.Recently also make great efforts multiple media functions to be incorporated into individual equipment (such as, Moxi (Moses) the PC of media center and execution Windows XP media center version) in.Although the each equipment in described equipment provides a bit convenient to the function of its execution, each equipment lacks the general and simple access to most of media.In addition, usually because making described equipment often spend hundreds of dollars, manufacture by the needs of expensive processing and/or local storage.In addition, the consumer electronics in described modern times conventionally consumes a large amount of electric power, even when idle, also consumes a large amount of electric power, and this is meaning, and it is more expensive and waste energy along with the time.For example, if consumer forgets, equipment is cut off or switch to different video input, this equipment may continue operation.In addition,, because neither one equipment is solution completely in the middle of described equipment, so the storehouse of other equipment in itself and family must be combined, this still leaves mouse nest formula line and many Long-distance Control to user.

In addition, when many newer taking the Internet when basic equipment is suitably worked, its conventionally provide updating currently form (with its originally may be available form compared with) media.For example, make video usually only make audio-visual-materials flow via the mobile equipment in the Internet, and can not make usually to follow DVD interactive " extra items " to flow, as " making " of video, game or director's comment.This is due to the fact that interactive material is often to make with specific format, and this specific format is intended to for process interactive particular device in this locality.For example, in DVD, HD-DVD and Blu-ray Disc, each has the specific interactive form of himself.Any home media equipment or local computer (its may through exploitation to support all popular format) be by needs tip property (sophistication) and flexibility to a certain degree, and it may be for too costliness and complicated of consumer's operation speech.

This problem is increased the weight of, if introducing format in the future after a while, local device may not have the hardware capabilities of supporting format, and this will mean that consumer will buy the local media equipment of upgrading.For example, for example, if (introduce the video of high-resolution or three-dimensional video-frequency on date after a while, video flowing of each eye), local device may not have the computing capability of this video of decoding, or it (for example may not have hardware for export this video with format, suppose that the 120fps video by synchronizeing with shading eyeglasses (shuttered glassess) realizes stereoscopic vision, wherein 60fps is sent to each eye, if consumer's vision hardware only can be supported 60fps video, this option is in the case of lacking the hardware purchase of upgrading unavailable).

In the time referring to most advanced and sophisticated interactive media (especially video-game), media device problem discarded and complexity is serious problems.

Modern video games application is divided into four main non-portable formula hardware platform: Sony substantially 1,2 and 3 (PS1, PS2, and PS3); Microsoft and Xbox and Nintendo Gamecube (Nintendo's sugar cube) and Wii ^tM; And taking PC as basic game.Each in described platform is different from its other, and the game that makes to be write to carry out on a platform can not carried out conventionally on another platform.Also may there is the consistency problem of generation equipment and equipment of future generation.Even if most software game developer sets up the software game that is independent of particular platform and design, in order to carry out particular game on particular platform, also need proprietary software layer (it is often called as " development of games engine " ") to adjust game to use on particular platform.Each platform is sold to consumer with the form of " control desk " (also, being attached to the off line box (standalone box) of TV or monitor/loud speaker) or itself is for PC.Conventionally, video-game is sold on the optical media such as blue-ray DVD, DVD-ROM or CD-ROM, and this optical media contains the video-game that is presented as most advanced and sophisticated real-time software application program.Along with family's broadband speeds increases, video-game just day by day becomes and can be used for downloading.

Because real-time and the high calculation requirement of advanced video game make realization extreme harsh with the particularity requirement of the platform compatibility of video game software.For example, a people from generation video-game to video-game of future generation (for example may expect, from XBox to XBox 360, or from Playstation 2 (" PS2 ") to Playstation 3 (" PS3 ")) complete game compatibility, as existing from a PC for example, to the general compatibility of productivity application program (, MicrosoftWord (Microsoft's word processor)) of another PC with faster treatment unit or core.But, be not this situation for video-game.Because in the time of the video-game of a distribution generation, video game manufacturers is sought the high likelihood energy for given price point conventionally, so often system is carried out to dynamic schema change, can not work in generation system after a while with the many game that make to be write for former generation system.For example, XBox is based on x86 series processors, and XBox 360 is based on PowerPC series.

Can utilize technology to imitate previous framework, but supposition video-game is real-time application, in imitation, reaches identical behavior usually unrealistic.This is the loss to consumer, video game console manufacturer and video game software publisher.For consumers, this means old generation video game console and new generation video game console maintenance are switched to TV to can play the necessity of all game.For control desk manufacturer, this means the cost relevant to the imitation of new control desk and slow employing.And for publisher, this means may must issue multiple versions of new game for example, to for each trade mark of video-game (contain all potential consumers-not only issue, XBox, Playstation) version, and usually distribution for example, for the version of each version (, PS2 and PS3) of given trade mark.For example, the independent version of " the mad rugby 08 " of exploitation Yi electricity Co., Ltd (Electronic Arts) is for XBox, XBox 360, PS2, PS3, Gamecube, Wii and PC platform and other platforms.

Portable set (such as, mobile phone and portable electronic device) also game developer is challenged.Day by day increase ground, described equipment connection is to radio data network and can foradownloaded video play.But, in market, there is multiple mobile phone and the media device with multiple different display resolution and computing capability.And, because described equipment has power consumption, cost and weight constraints conventionally, so it conventionally lacks the advanced figure accelerating hardware (such as, the equipment of being manufactured by the NVIDIA (the tall and handsome company that reaches) of the Santa Clara of California, USA) that is similar to Graphics Processing Unit (" GPU ").Therefore, merchants for game software development is developed the given game title for many dissimilar portable sets simultaneously conventionally.User can find: given game title is not useable for its specific mobile phone or portable electronic device.

Under the situation of family's game console, hardware platform manufacturer collects the royalty for issue the ability of game on its platform to software game developer conventionally.Mobile phone wireless communications carriers is also collected the royalty for game being downloaded to mobile phone to game publisher conventionally.Under the situation of PC game, there is not the royalty paying for issuing game, but serving burden compared with high consumption person that game developer is faced is conventionally expensive due to the installation question for supporting multiple PC configuration and may cause.And the piracy of the common less obstruction Games Software of PC, because it can and more easily be assigned with (for example,, via the Internet) more easily by piracy by user's reprogramming and the game of being proficient in technology easily.Therefore,, for software game developer, on game console, mobile phone and PC, distribution has cost and disadvantage.

For the game publisher of control desk and PC software, cost more than that.For via retail channel allocation game, publisher collects wholesale price lower than commercial value so that retailer has profit margin to retailer.Publisher also must pay the cost of manufacturing and distributing the physical medium of preserving game conventionally.Retailer often also to publisher collect " price protection expense " with contain possible accidental cost (such as, game can not sell, or game price reduce, or retailer must return some or all wholesale prices and/or regain game from buyer).In addition, retailer also collects the expense for helping to sell in flyer game to publisher conventionally.In addition, retailer day by day increases and buys back game from the user who has played game, and then described game is sold with used game, does not conventionally share the income of used game with game publisher.The following fact has increased the cost burden that imposes on game publisher: game is usually downloaded and freely copies for user via the Internet piracy and distribution.

Along with the Internet broadband speed increases and broadband connection becomes more extensively (more specifically, to family with to " Internet bar " of the PC of lease connecting Internet) in the U.S. and the whole world, game is assigned to PC or control desk more via download.And broadband connection is more for playing many people and MMOG (both are referred to by initial " MMOG " in the disclosure for this).These changes have alleviated some costs and the problem relevant to retail distribution.Download some disadvantages that game on line has solved game publisher, because apportioning cost is conventionally less and exist less or do not have a cost of not selling media.But the game of having downloaded is still by piracy, and because making it may spend the very long time, its size (size is usually several gigabytes) downloads.In addition, multiple game can be filled minidisk driver, those disc drivers of for example selling together with portable computer or together with video game console.But, need on-line joining process to make with regard to the degree that can play of game with regard to game or MMOG, problem of piracy is alleviated, because conventionally need user to have effective user account.Be different from and can (for example record the linear media that copies from the audio frequency of loud speaker by the video of camera filmed display screen or by microphone, video and music), it is unique that each video-game is experienced, and can not copy by simple video/audio record.Therefore, even carry out in the Copyright Law and pirate rampant region not powerful, also can protect MMOG to avoid by piracy, thereby can support business.For example, successfully dispose " World of Warcraft " MMOG of Vivendi SA (Vivendi Universal Games company), and do not suffer piracy in the whole world.And many online or MMOG game (such as, " Second Life " MMOG of Linden Lab (woods steps on laboratory)) produce the income of game operator by building economic model in game in, wherein assets can be used online tool and bring, sell and even set up.Therefore, can come to pay for the use of game on line by the mechanism buying or subscribe to except traditional game software.

Although because the character of online or MMOG makes usually can alleviate piracy, game on line operator still faces all the other challenges.Many game need a large amount of this locality (also, in family) processing resources suitably to work for online or MMOG.For example, if user has the local computer (, not having the computer of GPU, such as low-end notebook computer) of low performance, it may not play this game.In addition, along with game console is aging, it far lags behind current state of the art and may not process more senior game.Even if supposition user's local PC can process the calculation requirement of game, usually also exist complexity is installed.May there is driver incompatibility (for example, if download new game, the graphdriver of redaction may be installed, it causes the previous mounted game that depends on legacy version graphdriver to operate).Along with downloading more game, control desk may be finished local disk space.For example, if carried out (revising in the time finding defect and repair or to game, if game developer finds the too difficult object for appreciation of the rank of game or too easily plays), complicated game receives the patch (patch) of download conventionally from game developer along with passage of time.This patch needs new download.But be not the download that all users complete all patches sometimes.Other time, the patch of download is introduced other compatibilities or disk space consumption problem.

And, during game play, may need large data to download so that figure or behavioural information are provided to local PC or control desk.For example, if user enters in a room in MMOG, and run into the scene or the personage that form or have disabled behavior on user's local machine by graph data, must download that scene or personage's data.If it is fast not that the Internet connects, this can cause the essence between game play session to postpone.In addition, if the scene running into or personage need to exceed the storage area of local PC or control desk or the storage area of computing capability or computing capability, it can produce following situation: wherein user can not continue in game, or the figure that must reduce with quality continues.Therefore, online or MMOG game usually limits its storage and/or computation complexity requirement.In addition, it usually limits the amount that the data during game transmit.Online or MMOG game also can make the user's that can play games market narrow.

In addition the user who, is proficient in technology carries out reverse engineering and revises game so that they can be practised fraud the local duplicate of game more and more.Cheating may be with to carry out than pressing with the possible fireballing repeat knob of manpower (for example,, for shooting very rapidly) equally simple.Supporting in game in the game of transaction in assets, cheating can reach the swindle degree that causes bogus transaction to relate to having the assets that real economy is worth.When online or MMOG economic model is during based on described transaction in assets, this can cause the harmful consequence of the essence of game operator.

The cost of exploitation new game for example, along with PC and control desk (can be made more and more most advanced and sophisticated game, have figure more true to nature (such as, ray tracing in real time), and behavior more true to nature (such as, real-time physical is learned very)) and increase.Early stage in video-game industry, video-game exploitation is very similarly process of application software exploitation; Also, most of development costs (compare with the exploitation of figure, audio frequency and behavioral primitive or " assets ") in the exploitation of software, such as those software developments that can be developed the film for having extensive special-effect.Now, the video-game development at many tips is more similar to rich specific film exploitation than software development.For example, many video-games provide the emulation in the 3-D world, and produce personage, stage property and the environment of more true (also, seeming and outdoor scene (live action) image the taken computer graphical equally true to nature of photographing).One in the tool challenge aspect of photo development of games equally true to nature is to create to be different from the face that the computer of outdoor scene face produces.Face capture technology (such as, by the Contour of the Mova exploitation of the San Francisco of California ^tM(profile ^tM) authenticity capture systems) catch performing artist's facial volume precise geometry and follow the trail of performing artist's facial precise geometry with high-resolution in the time that performing artist is in motion.This technology allows in PC or game console, to reproduce 3D face, and in fact this 3D face can not be different from caught outdoor scene face.Accurately catching and reproducing " photo is equally true to nature " face is useful in many aspects.First, highly discernible famous person or sportsman are usually for video-game (usually being employed by expensive), and imperfection may be apparent for user, thereby make viewing experience (viewing experience) divert one's attention or make us unhappy.Frequently, need high degree of detail to realize the same realism of photo of height-need potentially and reproduce a large amount of polygons and high-resolution texture (connect at frame at polygon and/or texture on the basis of frame along with face moves changing).

In the time thering is the high polygon counting scene of detailed texture and change fast, support the PC of game or game console may not there is enough RAM to store enough polygons and the data texturing of the requisite number object animation frame producing for the fragment of playing.In addition, conventionally can be used for single optical drive or single disc driver in PC or game console conventionally much slow than RAM, and conventionally can not catch up with GPU reproducing acceptable maximum data rate in polygon and texture.Current game is loaded into most of polygons and texture in RAM conventionally, this means given scenario in complexity and on the duration, is subject to a great extent the capacity limit of RAM.Under the situation that for example FA Facial Animation is made, this may be limited to the low resolution face that there is no the sense of reality by PC or game console, or is only limited to and can and be written into the sense of reality face that is made into before animation for the polygon of multiframe more and texture (and other data) at a limited number of frame in suspension of game play.

When PC or control desk show while being similar to the message of " being written into ... ", watch process bar across screen lentamente mobile being known as by the user of complicated video-game be now latent defect.Next scene from disk (unless had ready conditions in addition, otherwise " disk " herein refers to non-volatile optical media or magnetic medium, and such as the nonmagnetic disk media of semiconductor " flash memory " memory) delay while being written into can spend some seconds or even some minutes.This loses time and may make game player quite dejected.As discussed previously, a large amount of or all delays may be the time that is written into due to the polygon from disk, texture or other data, but may be also following situation: in the time that the processor in PC or control desk and/or GPU prepare the data for scene, a cost part is written into the time.For example, Association football video-game can allow player to select in the middle of a large amount of players, group, sports ground and weather condition.Therefore, depend on and select what particular combinations, may need different polygons, texture and other data (being referred to as " object ") (for example, different groups have different color and pattern on its uniform) for scene.Enumerate possibly many or all arrangements in various arrangements and in advance in advance many or all objects in calculating object and by object stores at the disk for stores games.But if the number of arranging is very large, the required storage capacity of all objects may be excessive so that can not be arranged on disk (or too unrealistic so that can not download).Therefore the time that is written into that, existing PC and control desk system are conventionally tied and are grown for complex scene in the complexity of given scenario and playing duration time.

Another of the video game system of prior art and application software system is significantly limited in: its use more and more for example large database concept of 3D object (such as, polygon and texture), described large database concept need to be loaded in PC or game console for processing.As mentioned above, in the time that described database is stored on disk in this locality, described database can spend for a long time and be written into.But, if data base set is stored in remote location and carrys out access via the Internet, be written into the time conventionally much serious.Under this kind of situation, download large database concept may spend a few minutes, several hours or even several days.In addition, described database usually produces a large amount of expenses (for example, for playing, the detailed high 3D model that has mast sailing boat of film or historical record sheet) and is intended to for being sold to local terminal user.But once database is downloaded to local user, it just has by pirate risk.Under many situations, user is only for example, for its needs that whether are applicable to user (are watched in assessment data storehouse, when user carries out specificly when mobile, whether there is satisfied outward appearance or appearance for the 3D clothes of game) object and wish downloading data storehouse.For assessing the user of 3D database before buying in decision, it may be to hinder that length is written into the time.

Similar Problems occurs in MMOG (more specifically, as allowed user to utilize the personage's who more customizes game).For the PC or the game console that show personage, its behavior that needs can access to have 3D geometry (polygon, texture etc.) and described personage (for example, if personage has shield, whether shield enough strong so that lance deflection) database.Conventionally, in the time that MMOG is played for the first time by user, can use under the initial duplicate of game for personage's mass data storehouse, the initial duplicate of game is available or be downloaded to disk on game optical disk in this locality.But, along with game progress, if user run into database in this locality disabled personage or object (for example, if another user has set up certain personage processed), before can showing this personage or object, must download its database.This can cause the essence of game to postpone.

The tip property of given video-game and complexity, be previously another challenge of video-game developer and publisher in technique treats frequency game console situation: exploitation video-game often spends 2 years to 3 years, and cost is at tens million of dollars.Suppose that new video game control desk platform is with the roughly speed introducing every twice-a-decade, game developer need to start the development of those game the several years before the distribution of new game control desk, to make video-game simultaneously available in the time of the new platform of distribution.Several control desks from competitive manufacturer (are for example issued sometimes approximately simultaneously, each other in 1 year or in two years), but remain to be seen is the popularity (for example, which control desk is sold producing maximum video game software) of each control desk.For example, in the nearest control desk cycle, Microsoft XBox 360, SonyPlaystation 3 and Nintendo Wii plan to introduce in the approximately identical cardinal principle period.But in the several years before described introduction, game developer in fact must " pressure injection (place bets) " which control desk platform will be more successful than its other, and correspondingly drop into it and exploit natural resources.Studio also must estimate the successful film of possibility and share its limited making resource based on it for a long time before film distributor.The growth degree of the required investment of given video-game, game making all the more becomes similar film making, and game making company conventionally based on it to successfully estimating the future of particular video frequency game and dropping into it and make resource.But, be different from film company, the not only success based on making itself of this pressure injection; But it is dependent on the success of the game console that game will carry out thereon.On multiple control desks, distribution game can alleviate risk simultaneously, but this extra effort increases cost, and often postpones the actual distribution of game.

It is upper intensive, dynamic and interactive that application software on PC and user environment are just becoming more calculating, not only makes it visually more attract user, and make it more useful and directly perceived.For example, new Windows Vista (form distant view) ^tMoperating system with both have been incorporated to vision animation effect the later release of operating system.Advanced figure instrument (such as, from the Maya of Autodesk (Ou Teke) company ^tM(Mayan ^tM)) provide very most advanced and sophisticated 3D to reproduce and cartoon making ability (it has promoted the restriction of CPU and the GPU of current state of the art).But the calculation requirement of these new tools has produced many practical problems for the user of described product and software developer.

Because the vision of operating system (OS) show must be in the upper work of multiple computer (comprising the former generation computer of not resaling but still can upgrade along with new OS), OS graphics request is subject to the minimum common ground restriction of the computer (it generally includes the computer that does not comprise GPU) that OS will be used for to a great extent.This seriously limits the graph ability of OS.In addition, battery powered portable computer (for example, notebook) restriction vision display capabilities, because the high computational activity in CPU or GPU causes higher power consumption and shorter battery life conventionally.Portable computer is usually included in the unfavorable processor activity that automatically lowers during with processor to reduce the software of power consumption.In some computer model, user can manually be lowered processor activity.For example, the VGN-SZ280P notebook of Sony is included in a side and is labeled as " Stamina (persistence) " (for low performance, longer battery life) and opposite side on be labeled as the interchanger of " Speed (speed) " (for high-performance, shorter battery life).Even if the OS carrying out on portable computer also can work by land used must carrying out with the sub-fraction of its peak performance ability at computer.Therefore, OS graphics performance usually remains the available computing capability far below current state of the art.

Often sell intensive application program (as Maya) in high-end calculating, expect that described application program will be used for power PC.This conventionally produces much higher performance, and the requirement of more expensive and portable poor, minimum common ground.Therefore, described application program have than general purpose O S (or general productivity application program, similar Microsoft Office) limited many target audience and conventionally to sell than general purpose O S software or much lower the measuring of general purpose application program software.Potential audient is further restricted, because the user of expection is difficult to try out in advance application program intensive in described calculating often.For example, suppose that student wishes to understand the potential buyer who how to use Maya or known described application program and in purchase, wishes Maya on probation before investing (this may relate to also purchase can carry out Maya high-end computer).In the time that student or potential buyer can download the crippled version of Maya or obtain the physical medium duplicate of Maya crippled version, for example, if whole potential that its shortage can be carried out Maya (, process complicated 3D scene) computer, it can not carry out the comprehensive assessment of product.This limits in fact the audient of described high-end applications program.This also makes commercial value uprise, because development cost is conventionally through sharing than the much smaller purchase number of times of the purchase number of times of general purpose application program.

High price application program also swashs using the individuality of pirate duplicate of application software and business to produce more thorniness.Therefore, high-end applications program software suffers rampant pirate, although the publisher of this software has carried out making great efforts in a large number to alleviate this piracy by various technology.But even in the time using pirate high-end applications program, the PC that user also can not get rid of the current state of the art of investing costliness carries out the needs of pirate duplicate.Therefore,, although user can obtain by the sub-fraction of the actual retail price of software application the use of software application, the user of piracy software still needs buy or obtain expensive PC, to utilize this application program completely.

This user for the game of high-performance pirate video sets up equally.Although bootlegger can be played by the sub-fraction of the real price of game, it still needs to buy the required expensive computing hardware (for example, GPU-in-dash computer, or the high-end video game console of similar XBox 360) of suitably playing games.Supposing normally consumer's amusement of video-game, may be too expensive for the extra cost of high-end video game system.This situation is for example, even worse in the country (, China) of current workman's average year income quite low (with respect to current workman's average year income of the U.S.).The population of like this, much smaller percentage has high-end video game system or high-end PC.In these countries, the payable expense of user is to be used " Internet bar " of the computer that is connected to the Internet quite general.Frequently, described Internet bar have performance characteristics (such as, originally can make player can play the GPU that calculates upper intensive video-game) compared with old model or low side PC.This is the successful key factor of game carried out on low side PC (such as, " World of Warcraft " of Vivendi, it is highly successful in China, and normally plays in Chinese Internet bar).By contrast, in calculating, on the intensive more impossible PC being installed in Chinese Internet bar of game (as " Second Life "), play.In fact described game is inaccessible for the user of the low performance PC in only can access Internet bar.

Buy video-game and be first ready that the user of the demo version of trial game also exists obstacle by demonstration being downloaded to its family via the Internet for considering.Video game presentation is usually the all-round version of game, and some of them feature is stopped using, or the amount of game play is applied to restriction.This may relate to the growth process (possibility several hours) of downloading the data of billions of bytes on PC or control desk and on PC or control desk before execution game being installed on.Under the situation of PC, it also may relate to and calculates game and need which special driver (for example, DirectX or OpenGL driver), downloads correct version, correct version is installed, and is then determined whether PC can play this game.The latter's step may relate to determines whether PC has enough processing (CPU and GPU) ability, enough RAM and compatible OS (for example, some game carry out on Windows XP and on Vista, do not carry out).Therefore,, after attempting to carry out the growth process of video game presentation, the PC configuration that user may find in the case of given user, video game presentation can not be played.Even worse ground, once user has downloaded new driver for attempting this demonstration, these driver versions just may on PC, be accustomed to user use other game or application program incompatible, therefore, the installation of demonstration can cause previous exercisable game or application program not to operate.These obstacles not only make user dejected, and it also sells its game generation obstacle to video game software publisher and video-game developer.

Another problem that causes not having economic benefit and the following fact about: given PC or game console are designed to the performance requirement of the specific degrees that adapts to application programs and/or game conventionally.For example, some PC have RAM more or less, more slowly or faster CPU and GPU (if it has GPU) more slowly or faster.Some game or application program are utilized the full computing capability of given PC or control desk, and some game or application program are not utilized the full computing capability of given PC or control desk.If the selection of user's game or application program does not reach the peak performance ability of local PC or control desk, user may waste financial resources due to the feature of not utilizing on PC or control desk.Under the situation of control desk, control desk manufacturer may pay than subsidizing the desired many costs of control desk cost.

Be present in video-game sale and enjoy in another problem relate to user implement buy game before allow user to watch other people to play games.There are the some prior art approach to reset in time after a while for recording of video game.For example, United States Patent (USP) the 5th, instruct for 558, No. 339 during " game play " game state information (comprising game console action) has been recorded in video-game client computer (being had by same or different user).This state information can be in time use after a while with the some or all of game actions of for example, above resetting at video-game client computer (, PC or control desk).The remarkable shortcoming of the method is: for the user who watches the game of having recorded, user must have and can play the video-game client computer of this game and must have the video game application programs of carrying out on this computer, so that game play is identical when the game state that proper playback is recorded.In addition, video game application programs must be not exist the mode of possible execution difference to write with the game being recorded and between the game of playback.

For example, game graph substantially connects on frame basis and calculates at frame.For many game, depend on scene whether complicated especially or no existence slow down execution other (for example postpone, on PC, another process may be carried out, so that seize cpu cycle from game application), game logic may spend that shorter than a frame time or longer than the frame time time is calculated as next frame and the figure that shows sometimes.In this kind of game, " threshold value " frame for example, calculating with the time slightly fewer than a frame time (, few several cpu clock cycle) finally can occur.In the time using identical game state information again to calculate this Same Scene, may easily spend than the time in many several cpu clock cycles of frame time (for example, if internal cpu bus a little with not homophase of outside DRAM bus, even and there is not the large delay that seizes another process of several milliseconds of CPU times since game processing, it also introduces the delay of several times cpu cycle).Therefore,, in the time that playback is played, frame becomes with two frame times calculating but not calculates with single frame time.The frequency (for example, when game sampling from the input of game console time) of new frame is calculated in some behaviors based on game.In the time of game replaying, can not affect game play for this deviation of the time reference of different behaviors, but it can cause the game of institute's playback to produce Different Results.For example, if the track of basketball is to calculate with stable 60fps speed, but the speed that game console input is the frame based on calculating samples, in the time of recording game, the speed of the frame calculating may be 53fps, and in the time resetting game, the speed of the frame of calculating may be 52fps, whether this can make basketball be prevented from entering in basket to there are differences, thereby causes Different Results.Therefore, use the game of game state recording of video to need very careful Games Software design, to guarantee that using same game state information to reset produces identical result.

The video output of only recording PC or video game system (for example,, to VCR, DVD register, or to the video capture board on PC) for another prior art approach of recording of video game.Then video can be returned down and reset, or alternatively, the video of record be uploaded to the Internet (conventionally after by video compression).The disadvantage of the method is: in the time of playback 3D game sequence, user is only limited to from watching point (sequence from this watch be a little recorded) to watch sequence.In other words, watching a little of the immutable scene of user.

In addition, the compressed video of the game sequence of the record of playing on via the Internet PC of Er Shi family or game console, for other users are when available, even if video is Real Time Compression, also can not upload to the Internet by compressed video in real time.Its reason is to have highly asymmetric broadband connection (for example, DSL and cable modem have the dirty bandwidth more much higher than upper reaches bandwidth conventionally) because be connected in the world many families of the Internet.Compressed high-resolution video sequence usually has the bandwidth higher than the uploading bandwidth capacity of network, and it can not be uploaded in real time.Therefore,, after game replaying sequence (possible a few minutes or even several hours), before another user on the internet can watch this game, will exist significantly and postpone.Although this delay under particular condition (for example, watch the game player's who occurs at previous time achievement) tolerable, but it has eliminated for example, " at once resetting " ability when watching the ability of game on-the-spot broadcasting (the basketball championship of, being played by winning player) or broadcasting live ground game replaying.

Another prior art approach allows the beholder with television receiver to watch video-game on-the-spot broadcasting, but only under TV programme producers' control.Some television channels in the U.S. and other countries provide video-game to watch channel, and wherein televiewer can watch specific users of video games (for example, participate in the tournament annoying top player) on video-game channel.This is by being fed to the video output of video game system (PC and/or control desk) for video distribution and the treatment facility of television channel and completing.This situation during as the live Basketball Match of television channel broadcaster, wherein several cameras provide live from basketball court different angles around and are fed to.Television channel then can utilize its video/audio to process and effect device operates the output from various video game systems.For example, can superpose on the video from the video-game different players' of instruction the word (word superposes during can live at the scene Basketball Match as it) of state of television channel, and television channel can add the audio frequency of record from commentator's (it can discuss the action occurring between match period).In addition, can for example, by the camera of the video of the actual player of video-game output and recording game (, showing the emotional reactions of player to game) combination.

A problem of the method is: must make in real time video distribution and treatment facility that described live video feed is television channel use, to make it have the excitant of live broadcast.But as discussed previously, in the time that video game system is carried out from family (especially when a part for broadcast comprises when just catching game player's the live video of camera of real world videos), this is usually impossible.In addition, under championship situation, what pay close attention to is that in family, player can revise game and cheating, as discussed previously.Due to these reasons, described video-game broadcast on television channel is usually disposed and (is for example gathered in public location, at television stadio place or in arena) player and video game system, wherein TV making equipment can be accepted the video feed from multiple video game systems and potential live camera.

Although described prior art video-game television channel can be televiewer provide very the performance that stimulates (this be similar with live motion event (for example, similar with the video gamer presenting with " sportsman ") experience, not only action in the video-game world according to it, and action in real world according to it), but these video game systems are usually limited to player's hand-to-hand situation on health each other.In addition,, because television channel is broadcasted, the channel being broadcasted so each only can show a video flowing of being selected by the producer of television channel.Expensive due to these restrictions and airtime, making apparatus and producer, described television channel only shows the top player who participates in top championship conventionally.

The given television channel of in addition, broadcasting the full screen image of video-game to whole televiewers only shows a video-game at every turn.This seriously limits TV viewer's selection.For example, TV viewer may lose interest in to the game that show preset time.Another beholder may be only not interested in the game play of the particular player of projection preset time by television channel watching.Under other situations, beholder is may be only interested in watching expert player how to process specific rank in game.Other beholders may wish control watch point (video-game is watched a little from this), this is watched and is a little different from watching a little by selections such as production groups.In brief, TV viewer may have countless preference (even if several different television channels can be used, the specific broadcast of TV network is also not suitable with described preference) watching in video-game.Due to all above-mentioned reasons, make prior art video-game television channel in presenting video-game to TV viewer, there is remarkable restriction.

Another shortcoming of prior art video game system and application software system is: they are very complicated, and conventionally suffers mistake, collapse and/or unconscious and unwanted behavior (being referred to as " defect ").Although conventionally experience debug and tuning process (are often called " software quality assurance " or SQA) before distribution for game and application program, but almost constant: once game or application program are issued to the numerous audients in field, defect will occur suddenly.Regrettably, software developer is difficult to identify and track many defects after distribution.Software developer may be difficult to recognize defect.Even in the time that it understands defect, also may only exist it to can be used for identification is the limited amount information what causes this defect.For example, user can phone consumer's service hotline of game developer and leave message, this message statement: in the time playing games, screen starts flicker, then becomes solid blue (solid blue) and PC and freezes.It provides following the trail of useful considerably less information in defect for SQA group.Some game of on-line joining process or application program sometimes can provide more information under particular condition.For example, sometimes can use " house dog " process monitors whether " collapse " of game or application program.House dog process can collect game or when application crash about game or application program status of processes (for example, memory stack uses the degree that state, game or application program proceed to etc.) statistics, and then via the Internet, described information is uploaded to SQA group.But in complexity game or application program, this information can spend the very long time and decipher, to determine exactly what user carrying out in the time of collapse.However, also can not determine that what sequence causes collapse.

The another problem being associated with PC and game console is: it stands the service problem that makes consumer extremely not convenient.Service problem also affects the manufacturer of PC or game console, because it need to send special boxes conventionally to load and transport safely damaged PC or control desk, and thereby causes the cost (if PC or control desk are within guarantee period) of repairing.Lost on sale (or the online service use) impact that game or application software publisher are caused by PC in repairing state and/or control desk also can.

Fig. 1 illustrates such as Sony 3, Microsoft Xbox nintendoWii ^tM, taking Windows as basic personal computer or the prior art video game system of Apple Macintosh.Each in described system comprises CPU (CPU) for program code (being generally the Graphics Processing Unit (GPU) for carrying out advanced figure operation), and for the I/O (I/O) of multiple forms of external equipment and telex network.For the sake of simplicity, described assembly is shown as and is combined as individual unit 100.The prior art video game system of Fig. 1 is also shown as and comprises optical media driver 104 (for example, DVD-ROM driver); For storing the hard disk drive 103 of video game program code and data; For playing multiplayer, connecting 105 for the network of download games, patch, demonstration or other media; For storing the random access memory (RAM) 101 of the current procedure code of just being carried out by CPU/GPU 100; For receive the game console 106 from user's input command during game play; And display device 102 (for example, SDTV/HDTV or computer monitor).

In Fig. 1, shown prior art system is subject to some restrictions.First, compare with the access speed of RAM 101, optical drive 104 and Hard Disk Drive 103 often have much slow access speed.In the time directly working by RAM 101, because RAM 101 conventionally has much higher bandwidth and can not be subject to the fact that search that disk mechanism grows relatively postpones, CPU/GPU 100 can process the much more per second polygon figurate number of per second polygon figurate number possible when directly from hard disk drive 103 or optical drive 104 read routine codes and data in practice.But only limited amount RAM is provided in (for example, 256-512 Mbytes) in these prior art system.Therefore, usually need " being written into ... " sequence, wherein RAM 101 is periodically filled with the data for the next scene of video-game.

Some systems being written into and game play of program code of attempting side by side to overlap, but this only can (for example carry out in the time there is the event of known array, if just along driving path car, can in steered vehicle, be written into the geometry of the just approaching building in roadside).Change for complexity and/or rapid scene, the overlapping of this type is conventionally inoperative.For example, user, among campaign is carried out and in the view of carving at that time, RAM 101 is filled under the situation of data of indicated object completely, if user is moved to the left to watch the object in the current RAM of not being loaded in 101 rapidly by view, by causing the discontinuity of action, because do not exist time enough that new object is loaded in RAM 101 from hard disk drive 103 or optical media 104.

Another problem of the system of Fig. 1 is because the restriction of the storage volume of hard disk drive 103 and optical media 104 causes.For example, although Disk Storage Device can be manufactured into relatively large storage volume (, 50,000,000,000 bytes or more than 50,000,000,000 bytes), it is not still provided for enough storage volumes of the particular case running in current video game.For example, as discussed previously, Association football video-game can allow user to select in the middle of global many groups, player and sports ground.For each group, each player and each sports ground, need a large amount of texture mapping and environment mappings to come characterization 3D surface (for example, each group has unique sport shirt, and each needs unique texture mapping) in the world.

For a technology that solves above-mentioned the latter's problem be: for game, once user has selected texture and environment mappings, just calculate in advance texture and environment mappings.This can relate to process intensive in much calculating, comprises decompressing image, 3D mapping, adds shade, organising data structure etc.Therefore,, in the time that these calculating are carried out in video-game, may there is delay for user.Reduce on a methodological principle of this delay and be: each arrangement of carrying out all these when initial game development and calculating-comprise group, player's register and sports ground.The release version of game thereby will comprise all described through pretreated data on the one or more servers that are stored on optical media 104 or on the Internet, in the time that user makes one's options, only via the Internet, the selected pretreated data of selecting for given group, player's register, sports ground are downloaded to hard disk drive 103.But as practical problem, in game play, these data that are written in advance of possible each arrangement may be the data of several terabytes (terabyte) easily, it far exceedes the capacity of optical media equipment now.In addition the data of selecting for given group, player's register, sports ground, may be the data of several hundred million bytes or data more than several hundred million bytes easily.For example, the in the situation that of home-network linkups (, 10Mbps), connecting 105 these data of download via network will be than in the cost of local computing data for more time.

Therefore, in Fig. 1, shown prior art game structure makes user stand remarkable delay between the changing compared with large scene of complexity game.

Another problem of the prior art approach of all prior art approach as shown in Figure 1 is: in these years, video-game is tended to become more senior and needed more CPU/GPU disposal abilities.Therefore,, even if adopt endless RAM, video-game hardware requirement also exceedes the peak level of disposal ability available in described system.Therefore, need user to upgrade game hardware every several years to keep synchronous (or to play compared with new game compared with low quality level).Than the consequence of the trend of more senior video-game be in the past: conventionally do not have an economic benefit for the machine of the playing video game of domestic use, because its cost is determined by the requirement of its supported peak performance game conventionally.For example, may play with XBox 360 game of similar " war machine (Gears of War) ", this game requires the RAM of high performance CPU, GPU and several hundred million bytes, or may play with XBox 360 " eating beans (Pac Man) ", it is the game from 20 century 70s, and it only needs the RAM of several kilobytes and the unusual CPU of low performance.In fact, XBox 360 has enough computing capabilitys of " eating beans " game of many whiles of Co-location simultaneously.

In the great majority of a week hour, conventionally close video game machine.According in July, 2006 Nielsen (Nelson) amusement to the above research that enlivens player in 13 years old and 13 years old, on an average, enliven player spend in one week in 14 hours or one week whole hours only 12% play control desk video-game.This means that average video game console is idle within 88% time, this is the inefficiency use of expensive resource.Suppose that video game console is usually to subsidize to reduce purchasing price (expect that this subsidy will earn back by the royalty of buying from following video game software) by manufacturer, this is meaningful especially.

The cost that video game console also causes to almost any consumer electronics is relevant.For example, the electronic equipment of system and mechanism need to be contained in shell.Manufacturer need to provide service guarantees.The retailer who sells this system need to collect about the sale of system and/or about the profit of the sale of video game software.All of these factors taken together adds the cost of video game console, and this cost must be subsidized by manufacturer, be passed to consumer, or is subsidized by manufacturer and consumer.

In addition, piracy is the larger problem of video game industry.In fact the release mechanism utilizing on each larger video game system " breaks " in these years, causes the unwarranted of video-game to copy.For example, Xbox 360 safety systems were broken and user now can the illegal duplicate of download online in July, 2006.Downloadable game (for example, for PC or Mac game) especially easily stands piracy.In the specific region in the world (wherein pirate control is not strong), there is not in fact the feasible market of independent vide Games Software, because user can easily buy pirate duplicate with a very little part for cost with legal duplicate is general.And in many places in the world, the cost of game console accounts for the high percentage of income, even so that pirate controlled, also few people can afford the games system of current state of the art.

The income of video-game industry has been reduced in the market of the game having used in addition.When user becomes game while being weary of, it can be sold to game game is resell to other users' shop.Without permission this but general practice has significantly reduced the income of game publisher.Similarly, when existed platform to change every several years, conventionally occur that about 50% sale reduces.This be because: in the time that user knows the platform of forthcoming more recent version, user stops buying for for example, compared with the game of legacy platform (,, in the time forthcoming Playstation 3, user stops buying Playstation 2 and plays).Combine, the development cost of the loss of sale and the increase relevant to new platform can have to game developer's rentability the adverse effect of highly significant.

New game control desk is also very expensive.Xbox 360, Nintendo Wii and Sony Playstation 3 are all with hundreds of dollars of retails.The personal computer games system of high ability can spend up to 8000.This represents user's remarkable investment, specifically, considers that hardware stale and many systems after several years are the facts of buying for child.

A method of the problems referred to above is games on line, wherein games code and data host is managed on behalf of another on server and is sent on request client machine, and compressed video and audio frequency flow via digital broadband network.Some companies (such as the G-Cluster of, Finland (G-troop company), it is the subsidiary of Japanese SOFTBANK Broadmedia (the wide matchmaker of soft silver) now) are current provides described service online.It is available in local network (such as, those networks that provide in hotel and by DSL and cable TV supplier) that similar game services becomes.The larger shortcoming of these systems is problems of time delay, and also, signal rows enters game server and advances the spent time from game server, and game server is positioned in " front end " of operator conventionally.Between the time that quick acting video-game (also referred to as " very fast (twitch) " video-game) performs an action by game console user and the time of renewal display screen with the result of demonstration user action, need low-down time delay.Need low delay, feel that game " immediately " responds so that obtain user.The type of videogame and user's qualification and meet user with different delayed time interval.For example, for slowly unofficially for game (similar backgammon) or slow motion RPG, the time delay of 100 milliseconds may be tolerable, but in fast motion game, the time delay that exceedes 70 milliseconds or 80 milliseconds can cause user's performance more sorrily in game, and therefore unacceptable.For example, needing in the game of fast response time, when time delay is in the time being increased to 100 milliseconds for 50 milliseconds, exist the sharp of accuracy to fall.

In the time that game or application server are installed in the network environment of nigh controlled network environment or and/or tolerable bandwidth peak measurable to user's network path, aspect the consistency of maximum delay and time delay, control time delay much easier (for example, therefore user observes the stable motion mobile from digital video via network).The control of this degree can reached below: inter-household at cable tv network front end to cable tv user, or from DSL central office the family to DSL subscriber, or in commercial office's Local Area Network (LAN) environment from server or user.And the point-to-point private that likely obtains the particular hierarchical with guaranteed bandwidth and time delay between business connects.But game host is managed on behalf of another in the server-centric that is connected to common interconnection network and then make via broadband connection compressed video flow (stream) in user's game or application system, many factors cause time delay, cause the serious restriction in the deployment of prior art system.

In the typical family that connects broadband, user can have for the DSL of broadband service or cable modem.Described broadband service causes the nearly round trip time delay (and sometimes more) of 25 milliseconds between user's family and common interconnection network conventionally.In addition, there is the round trip time delay causing owing to routing data to server-centric via the Internet.The delay causing via the time delay of the Internet data when providing the route of data and data and be routed and changing.Except routing delay, also because light causes round trip time delay through the speed of the optical fiber that makes the interconnection of most of the Internets.For example, for every 1000 miles, because light causes the round trip time delay of approximately 22 milliseconds through speed and other consumptions of optical fiber.

Extra time delay can be owing to causing via the data rate of the mobile data in the Internet.For example, if user has the DSL service of selling with " 6Mbps DSL service ", in practice, user will probably obtain being at most less than the descending conveying capacity of 5Mbps, and may periodically see the connection degradation producing due to various factors (being written into crowded that time durations locates at Digital Subscriber Line Access Multiplexer (DSLAM) such as, peak value).If exist other places in crowded or cable modem system network to exist crowded in the local sharing coaxial cable via neighbor's circulation, Similar Problems can occur, thereby the data rate of the cable modem that is used for the connection of selling with " 6Mbps cable modem services " is decreased to much smaller than this data rate.If making the packet under the steady rate of 4Mbps flows via described connection from server-centric with User Datagram Protoco (UDP) (UDP) form uniaxially, if all are all suitably worked, packet will by and do not cause extra time delay, if but exist crowded (or other hinder) and only 3.5Mbps can be used for making data to flow to user, under typical situation, bag will be dropped, cause obliterated data, or grouping will be queued up until they can be sent out at crowded some place, thereby introduce extra time delay.Different crowded points have the different queue capacity for preserving the grouping being delayed, and therefore, under some situations, can not successfully solve immediately crowded grouping and abandon.Under other situations, by the data queue of millions of bits and its transmission the most at last.But under nearly all situation, the queuing at crowded some place has capacity limit, once and exceed this restriction, queue will overflow and divide into groups to be dropped.Therefore,, for fear of causing extra time delay (or even worse ground, packet loss), must avoid exceeding the data rate capacity from game or application server to user.

Also due to compressed video in server and in client device the required time of decompressed video cause time delay.In the time that next frame to be shown is being calculated in the video-game of carrying out, further cause time delay on server.Current available video compression algorithm is subject to high data rate or high time delay.For example, motion JPEG is the compression algorithm only damaging in frame, and this compression algorithm is characterized as low delay.Each frame of video is independent of each other frames of video and compresses.In the time that client device receives a frame of compressed motion JPEG video, its this frame and show this frame of can decompressing immediately, thus cause low-down time delay.But because each frame separately compresses, so algorithm can not utilize the similarity between successive frame, and therefore only in frame video compression algorithm be subject to very high Russia data rate.For example, 60fps (number of pictures per second) 640 × 480 motion JPEG videos may need 40Mbps (per second megabit) or the above data of 40Mbps (per second megabit).To be too expensive (and the way it goes for the application program based on the Internet of most consumers) for the described high data rate of described low-resolution video window in many broadband application programs.In addition, because each frame is through independent compression, so the false shadow in the frame that may produce due to lossy compression method may come across the diverse location place in successive frame.This can cause in the time of decompressed video, is mobile visual artifacts In the view of beholder.

Other compression algorithms (such as, from the MPEG2 of Microsoft company, H.264 or VC9) when the configuration for prior art, can realize high compression ratio, but taking high time delay as cost.Described algorithm utilizes compression in interframe compression and frame.Periodically, described algorithm is carried out the interior compression of only frame of frame.This frame is called as key frame (so-called " I " frame).Then, this algorithm is compared with previous frame I frame conventionally with successive frames.Not compress independently previous frame and successive frames, but algorithm is determined image from I frame to previous frame and successive frames has anything to change, and then this change is saved as: " B " frame (for the change before I frame) and " P " frame (for the change after I frame).This causes than only compressing much lower data rate in frame.But it is conventionally taking higher time delay as cost.I frame is conventionally than B frame or P frame much bigger (usually large 10 times), and therefore, spends pro rata the longer time with data-oriented speed rates.

Consider (for example) i situation: wherein I frame is the big or small 10 times of B frame and P frame, and in each single I frame, has 29 B frame+30 P frame=59 interframe, or for each " frame group " (GOP) 60 frames altogether.Therefore, under 60fps, 1 60 frame GOP of existence per second.Suppose that transmission channel has the maximum data rate of 2Mbps.In order to reach the video of first water in channel, compression algorithm will produce 2Mbps data flow, and given above-mentioned ratio, and this will produce 2 megabits of (Mb)/(59+10)=30 in every frame, 394 bits and 303,935 bits of every I frame.In the time receiving compressed video flowing by decompression algorithm, for displaying video stably, need to for example, decompress and show each frame with regular interval (, 60fps).In order to realize this result, if any frame is subject to transmission delay, need at least this time delay of all frame delay, therefore the frame time delay of situation will be defined for the time delay of each frame of video worst.Because I frame maximum, so I frame is introduced longer transmission time delay, and whole I frame must can decompress and show that I frame any interframe of I frame (or depend on) receives before.Suppose that channel data rates is 2Mbps, transmits I frame and will spend 303,935/2Mb=145 millisecond.

Use the inter-frame video compressibility (as mentioned above) of large percentage of the bandwidth of transmission channel by because I frame stands long delay with respect to the large size of the mean size of frame.Or, in other words, for example, when prior art interframe compression algorithm than the low average every frame data speed of compression algorithm in frame only (is reached, 2Mbps is to 40Mbps) time, it still suffers the every frame data speed of high peak value (for example due to large I frame, 303,935*60=18.2Mbps).But please remember: above-mentioned analysis supposition P frame and B frame are all much smaller than I frame.Although this sets up substantially, for having and the frame of previous frame, high motion or the incoherent hi-vision complexity of scene change, this is false.Under described situation, P frame or B frame can become and generally large (if P frame or B frame become larger than I frame, most advanced and sophisticated compression algorithm is replaced P frame or B frame by " pressure " I frame and with I frame conventionally) of I frame.Therefore, the data rate peak value of I frame sign can come across in digital video frequency flow at any time.Therefore, for compressed video, in the time that average video data rate approaches the data rate capacity of transmission channel (often for this situation, the given high data rate for video requires), cause high frame time delay from I frame or large P frame or the peak data rates of B frame.

Certainly the above-mentioned discussion compression algorithm time delay that only characterization is produced by the large B frame in GOP, P frame or I frame.If use B frame, time delay will be higher.Reason is because before can showing B frame, must receive B frame all B frames and I frame afterwards.Therefore, in picture group (GOP) sequence such as BBBBBIPPPPPBBBBBIPPPPP, wherein before each I frame, there are 5 B frames, only after receiving B frame subsequently and I frame, just can show a B frame by video decompressor.Therefore, if video is flowed with 60fps (also, 16.67 milliseconds/frame), before the B frame that can decompress, no matter how fast channel width is, receive five B frames and I frame and will spend 16.67*6=100 millisecond, and this is 5 B frame Russia situations only.The compressed video sequence with 30 B frames is quite general.In addition,, under as the low channel width of 2Mbps, the delay causing due to the size of I frame is increased to a great extent owing to waiting for that B frame arrives the delay producing.Therefore, on 2Mbps channel, the in the situation that of a large amount of B frame, use prior art video compression technology exceed 500 milliseconds or 500 milliseconds of above time delays quite easy.If do not use B frame (for given quality level, taking compared with low compression ratio rate as cost), do not cause the time delay of B frame, but still cause the time delay causing due to peak value frame sign as described above.

Problem is exactly because the character of many video-games increases the weight of.Utilize the video compression algorithm of gop structure as described above to be optimized to a great extent for using together with live video or the film material that will be used for passive viewing.Conventionally, camera (true camera, or the virtual camera under the situation of animation that computer produces) and scene relatively stable, if only because camera or scene move around very much with jolting, video or film material (a) are conventionally watched and are made us unhappy, if and (b) it is just viewed, in the time that camera jolts suddenly back and forth, beholder can not (for example closely follow this action conventionally, if camera in the time taking the child who blows out the candle on birthday cake by disturbance and jolt back and forth between cake suddenly, beholder concentrates on child and cake conventionally, and brief interruption when ignoring camera and moving suddenly).Under the situation of video talks or video conference call, camera can be immobilizated in fixed position and not and move, thereby cause at all considerably less data peaks.But 3D is high, action video game (is for example characterized by steady motion, consider 3D contest, wherein whole frame, in the duration of contest in rapid movement, or is considered first person shooting game, and wherein virtual camera moves around consistently with jolting).Described video-game can produce has the large and frame sequence of peak value frequently, and wherein user may need clearly to see between this unexpected moving period what has occurred.Therefore,, in the high action video game of 3D, compress the far not tolerable of false shadow.Therefore, the video of many video-games output (due to its character) produces and has the very high and compressed video flowing of peak value frequently.

The user who supposes fast motion video-game has little tolerance for high time delay, and given all above-mentioned time delay reasons, has so far the restriction for the video-game that server host that video flows is on the internet managed on behalf of another.In addition, if desired the application program of high interaction by Co-location on common interconnection network and make video flow, the user of described application program suffers similar restriction.Described service needs network configuration, wherein Co-location server is directly arranged in front end (under the situation of cable broadband) or central office (under the situation of Digital Subscriber Line (DSL)), or in LAN in business contexts (or the private of special classification connects), to control from client device the route to server and distance to minimize time delay and can adapt to peak value and do not cause time delay.LAN (being conventionally rated at 100Mbps-1Gbps) and the leased line with enough bandwidth can support peak bandwidth to require (sub-fraction that for example, 18Mbps peak bandwidth is 100Mbps LAN capacity) conventionally.

If carry out idioadaptation, peak bandwidth requires also can be adapted to by house broadband architecture.For example, in cable tv system, can be Digital Video Communication and provide dedicated bandwidth, this dedicated bandwidth can be processed the peak value such as large I frame.In addition, in DSL system, can supply the DSL modulator-demodulator (consideration peak value) of fair speed, maybe can supply the connection of the special classification that can process higher data rate.But the traditional cable modulator-demodulator and the DSL architecture that are attached to common interconnection network far can't stand for the peak bandwidth of the video for compressing requires.Therefore, online service (is managed on behalf of another video-game or application host in the server-centric apart from the long distance of client device, and then make the compressed video output streams moving via the Internet via traditional house broadband connection) suffer significant time delay and peak bandwidth requirement-especially for the game of the low-down time delay of needs and application program (for example, first person shooting game and other multi-users, interactive action game, or need the application program of fast-response time).

Brief description of the drawings

According to detailed description below and can more completely understand with reference to the accompanying drawings the disclosure, but, described accompanying drawing be not used for by disclosed subject matter restricted to shown in specific implementations, and just with laying down a definition and understanding.

Fig. 1 shows the framework of prior art video game system.

Fig. 2 a to Fig. 2 b shows according to the high level system architecture of an embodiment.

Fig. 3 shows for actual, the specified and required data rate of communicating by letter between client and server.

Fig. 4 a shows the host services and the client that use according to an embodiment.

Fig. 4 b shows the relevant exemplary time delay of communicating by letter between client and host services.

Fig. 4 c shows according to the client device of an embodiment.

Fig. 4 d shows the client device according to another embodiment.

Fig. 4 e shows the example block of the client device in Fig. 4 c.

Fig. 4 f shows the example block of the client device in Fig. 4 d.

Fig. 5 shows an example of the video compression that can use according to an embodiment.

Fig. 6 a shows an example of the video compression that can use in another embodiment.

Fig. 6 b shows and the peak value transmitting in the data rate of low complex degree, low action video Serial relation.

Fig. 6 c shows and the peak value transmitting in the data rate of high complexity, high action video Serial relation.

Fig. 7 a to Fig. 7 b shows the instance video compress technique using in one embodiment.

Fig. 8 shows the additional examples video compression technology using in one embodiment.

Fig. 9 a to Fig. 9 c shows the case technology using in one embodiment for alleviating data rate peak value.

Figure 10 a to Figure 10 b shows image image block is packaged in to an embodiment in grouping effectively.

Figure 11 a to Figure 11 d shows the embodiment that uses forward error correction technique.

Figure 12 shows an embodiment who compresses with multi-core processing unit.

Figure 13 a to Figure 13 b shows according to the geo-location between the host services of various embodiment and communication.

Figure 14 shows the relevant exemplary time delay of communicating by letter between client and host services.

Figure 15 shows example host services server-centric framework.

Figure 16 shows the example screen shot of an embodiment of the user interface that comprises multiple live video windows.

Figure 17 shows the user interface of the Figure 16 after selecting particular video frequency window.

Figure 18 shows the user interface of the Figure 17 after particular video frequency window is amplified to full screen size.

Figure 19 shows the example partner user video data on the screen that is superimposed upon multiplayer.

Figure 20 shows the instance user page for the game player in host services.

Figure 21 shows example 3D interactive type advertisement.

Figure 22 shows the case step sequence for catch the image that produces the photo realistic with grain surface from the surface of live performance.

Figure 23 shows the instance user interface page that allows to select linear media content.

Figure 24 shows the time quantum that disappears before live webpage at the scene and the curve chart of connection speed.

Embodiment

Enumerated in the following description specific detail (such as, device type, system configuration, communication means etc.), to provide thorough understanding of the present disclosure.But, generally it be to be appreciated that those skilled in the art that putting into practice described described embodiment may not need these specific detail.

Fig. 2 a to Fig. 2 b provides the high level architecture of two embodiment, wherein video-game and software application (are noted by customer site 211 by host services 210 Co-locations and under subscription service, " customer site " meaning is the position anywhere that user locates, if use mobile device to comprise outdoor) client device 205 located carrys out access via the Internet 206 (or other public networks or private network).Client device 205 can be have to the wired or wireless connection of the Internet, have inside or external display device 222 all-purpose computer (such as, macintosh computer taking Microsoft Windows or Linux as basic PC or Apple company), or it can be the private client equipment such as Set Top Box (having the wired or wireless connection to the Internet) that video and audio frequency is outputed to monitor or television machine 222, or it can be the action equipment presumably having to the wireless connections of the Internet.

The user input device that any one in described equipment can have himself (for example, keyboard, button, touch screen, tracking plate (track pad) or inertia sensing rod (inertial-sensing wand), Video Capture camera and/or motion tracking camera etc.), or it can use the external input device 221 (for example, keyboard, mouse, game console, inertia sensing rod, Video Capture camera and/or motion tracking camera etc.) by wired or wireless connection.As being below described in more detail, host services 210 comprises the server (comprising those servers with high ability CPU/GPU disposal ability) of various performance levels.At game replaying or during using the application program in host services 210, family or office's client device 205 receive from user's keyboard and/or controller input, and then it transfers to host services 210 by controller input via the Internet 206, host services 210 carry out games code as response and produce for play or the successive frames of the video output (sequence of video images) of application software (for example, if user presses the button that the personage that will guide on screen moves right, games then will produce the sequence of video images that shows that personage moves right).Then use low delay video compressor to compress this sequence of video images, and the then transmission low delay video flowing via the Internet 206 of host services 210.Family or office's client device then decode compressed video flowing and will through decompress video image be reproduced on monitor or TV.Therefore, reduce significantly calculating and the graphic hardware requirement of client device 205.Client 205 only needs to have for the input of keyboard/controller is forwarded to the Internet 206 and the disposal ability of the compressed video flowing receiving from the Internet 206 of decoding and decompress, in fact now any personal computer all can be on its CPU with software carry out these (the double-core CPU of Intel Company for example, carrying out with about 2GHz can decompress use such as H.264 and the 720p HDTV of the compressor reducer of windows media VC9 coding).In addition, under the situation of any client device, special chip also can than the much lower cost of universal cpu and than the power consumption of universal cpu much less (such as, modern PC is required) carry out in real time the video decompression for described standard.Notably, in order to carry out the function of the input of transmission controller and decompressed video, home client equipment 205 is without any need for specialized Graphics Processing Unit (GPU), optical drive or hard disk drive (such as, shown prior art video game system in Fig. 1).

Along with game and application software become more complicated and have more photo realistic sense, it will need CPU, the GPU of superior performance, more RAM, and disc driver more greatly and faster, and can make the computing capability at host services 210 places constantly upgrade, but terminal use will not need to make family or office's client platform 205 to upgrade, because the processing requirements that makes family or office's client platform 205 by given video decompression algorithm is kept to constant for display resolution and frame rate.Therefore, in Fig. 2 a to Fig. 2 b, in illustrated system, there is not being seen hardware constraints and consistency problem now.

In addition, because game and application software are only carried out in the server in host services 210, so in user's family or office (unless had ready conditions in addition, otherwise " office " will comprise any non-house background as used herein, comprise (for example) classroom) in never there is game or the duplicate of application software (form of optical media, or software for downloading).This significantly alleviate game or application software by the possibility of bootlegging (piracy), and alleviate can by play or application software use valuable database by pirate possibility.In fact, if desired the server of specialization (for example, need equipment very expensive, large or that have noise) play for family or office and use infeasible game or application software, even if obtain the pirate duplicate of game or application software, it also can not operate in family or office.

In one embodiment, to the design game of video-game or application software developer, (it refers to software supplier, game or film workshop to host services 210 substantially, or game or application software publisher) 220 provide SDK (Software Development Kit), to make it can design the game that can carry out in host services 210.Described instrument allows feature (described feature conventionally in independent PC or game console by unavailable) that developer utilizes host services (for example, the very large database (unless had ready conditions in addition, otherwise " geometry " will be used in reference to the polygon, texture, rigging, illumination, behavior and other assemblies and the parameter that limit 3D data set in this article) of fast access complex geometric shapes).

Under this framework, different business prototype are possible.Under a model, host services 210 from terminal use collect subscription charge with and pay royalty to developer 220, as shown in Fig. 2 a.Substitute implement in (shown in Fig. 2 b), developer 220 directly from user collect subscription charge with and pay the expense for Co-location game or application content to host services 210.Any particular business model for providing game on line or application host to manage on behalf of another is provided these general principles.

Compressed video properties

As discussed previously, provide online a prominent question of video-game service or application software service to be time delay.The time delay of 70 milliseconds-80 milliseconds (from input equipment by user activate (actuate) time be carved into the moment show response on display device time) for for needing the game of fast-response time and the upper limit of application program.But, because reality and physical constraint make to be very difficult to reach this framework shown in Fig. 2 a and Fig. 2 b in a large number.

As indicated in Figure 3, in the time that user subscribes to Internet service, connect the demarcation maximum data rate 301 of the family or the office that are conventionally rated for user.The strategy and the routing device ability that depend on supplier, this maximum data rate can more or less strictly be executed, but conventionally because the one in many different reasons makes actual available data rate lower.For example, may there is multi-network communication communication at DSL central office place or on local cable modem loop, or may on cable, there is noise, thus cause the grouping abandoning, or supplier may set up the monthly bit of maximum number of every user.Current, for the maximum downstream data rate of cable and DSL service conventionally in hundreds of kilobits/second (Kbps) in the scope of 30Mbps.Cellular services is limited to the downlink data of hundreds of Kbps conventionally.But the user's of the speed of broadband service and subscription broadband service number will sharply increase along with the time.Current, some analysts estimate that 33% U.S. broadband user has 2Mbps or downstream data rate more than 2Mbps.For example, some analysts prediction: to 2010 only, exceeding 85% U.S. broadband user will have 2Mbps or data rate more than 2Mbps.

As indicated in Figure 3, actual available maximum data speed 302 can fluctuate along with the time.Therefore,, in low delay, game on line or application software situation, be sometimes difficult to the actual available data rate of prediction for particular video stream.If for the scene complexity of specified quantitative and motion under the number of pictures per second (fps) of given number with given resolution (for example, 640 × 480@60fps) maintain the required data rate of given quality level 303 and raise higher than actual available maximum data speed 302 (as the peak value instruction by Fig. 3), can there are some problems.For example, some Internet services will only abandon grouping, thus cause the data of the loss on user's video screen and distortion/image lost.Other services will temporarily cushion (also, queue up) additional packet and with available data rate, described grouping will be provided to client, thereby cause the increase of time delay-be unacceptable result for many video-games and application program.Finally, some ISPs by the increase of data rate be considered as malicious attack (such as, deny service attack (by computer hacker with so that network connects inactive known technology)), and the Internet that will cut off user in the special time cycle connects.Therefore, embodiment described herein manages to guarantee that the desired data speed for video-game can not exceed maximum usable data rates.

Host services framework

Fig. 4 a illustrates according to the framework of the host services 210 of an embodiment.Host services 210 can be arranged in individual server in the heart, or can cross over multiple server-centrics and disperse (to think to have and provide low delay connection than the user in the path to particular server center of its other low delay, so that load balance to be provided between user, and provide redundancy under the out of order situation of one or more server-centric).Host services 210 finally can comprise thousands of or millions of servers 402 even, thereby serves very large user base (user base).Host services control system 401 provides the overall control to host services 210, and guides router, server, video compression system, charging and account system etc.In one embodiment, host services control system 401 is implemented on the distributed processing system based on Linux, and this treatment system is tied to for storing the RAID array for the database of user profile, server info and system statistics data.In the foregoing description, unless owing to other particular systems, otherwise the exercises of being implemented by host services 210 initial and control by host services control system 401.

Host services 210 comprises many servers 402, such as the current described server that can obtain from Intel (intel corporation), IBM (IBM) and Hewlett Packard (Hewlett-Packard) and the other thereof.Or, server 402 can be assembled into customizable component configuration, or finally server 402 can be integrated to be one single chip by whole server implementation.Although this figure shows minority server 402 for purposes of illustration, in actual deployment, may exist few to a server 402 or reach the server of millions of or millions of above servers 402.Server 402 all can configure in the same manner (as the example of some configuration parameters, has identical cpu type and performance; There is or do not have GPU, and if there is GPU, identical GPU type and performance there is; There is CPU and the GPU of similar number; There is the RAM of same amount and same type/speed; And there is identical RAM configuration), or each subset of server 402 (for example can have identical configuration, 25% server can an ad hoc fashion configure, 50% server configures by different way, and 25% server configures in another mode), or each server 402 can be different.

In one embodiment, server 402 is without disk, also be, (it is for optics or storage magnetic not to have himself the large capacity holder in this locality, or the holder of based semiconductor, such as other large capacity storage devices of flash memory or service similar functions), each server connects and the shared large capacity holder of access via quick base plate or network.In one embodiment, this is connected to the storage area network (SAN) 403 that is connected to raid-array (RAID) 405 series fast, between the equipment that uses ultrahigh speed Ethernet to implement, has connection.As known to persons skilled in the art, SAN 403 can be used for many RAID arrays 405 to combine, thereby causes high bandwidth-approaching bandwidth that can certainly obtain for the RAM of current game console and PC that maybe may exceed.In addition,, although the RAID array of the rotation media based on such as magnetic medium often has significant search time access time delay, the RAID array of based semiconductor holder can be embodied as has much lower access time delay.In another configuration, some or all of servers 402 provide some or all of himself large capacity holders in this locality.For example, server 402 can by the information of frequent access (such as, the duplicate of its operating system and video-game or application program) be stored on the holder based on low delay local flash memory, but it can utilize SAN to carry out the RAID array 405 with higher search time delay of access based on rotation media, with the large database concept of access geometry continually or game state information more not.

In addition, in one embodiment, host services 210 is used the low delay video compression logic 404 of below describing in detail.Video compression logic 404 can software, (below describing its specific embodiment) implemented in hardware or its any combination.Video compression logic 404 comprises the logic for compressed audio and visual material.

In operation, in the time of the application program at playing video game or 211 places, user place via keyboard, mouse, game console or other input equipments 421, the control signal 406a-b (being generally the form of UDP grouping) of the button press that represents to be impelled by user (and the user of other types input) is transferred to host services 210 by the control signal logic 413 in client 415.By be routed to from given user's control signal suitable server (if or multiple server in response to user's input equipment, route to multiple servers) 402.As illustrated in Fig. 4 a, can control signal 406a be routed to server 402 via SAN.Alternatively or in addition, can for example, via host services network (, the Local Area Network based on Ethernet) control signal 406b be routed directly to server 402.No matter how control signal 406a-b is transmitted, this or described server are all carried out game or application software in response to control signal 406a-b.Although undeclared in Fig. 4 a, but diverse network coupling assembling (such as, fire compartment wall and/or gateway) can processing host the edge (for example,, between host services 210 and the Internet 410) of service 210 and/or edge's (between the Internet 410 and family or office's client 415) of customer site 211 import into and spread out of communicate by letter.Performed game or the figure of application software and audio frequency output (are also, new sequence of video images) provide to low delay video compression logic 404, low delay video compression logic 404 according to low delay video compression technology (such as, described technology described herein) and compressed video image sequence and via the Internet 410 (or, as described below, via the optimum high speed network service of walking around common interconnection network) and compressed video flowing (conventionally having the audio frequency of compressed or uncompressed) transmission is back to client 415.Then, low delay video decompression logic 412 decompressed video in client 415 and audio stream also reproduce the video flowing through decompressing, and conventionally on display device 422, play the audio stream through decompressing.Or, can be on the loud speaker separating with display device 422 audio plays or audio plays not.Note, although input equipment 421 and display device 422 are shown as self contained facility in Fig. 2 a and Fig. 2 b, it can be integrated in the client device such as portable computer or action equipment.

Family or office's client 415 (previously in Fig. 2 a and Fig. 2 b, being described as family or office's client 205) can be the equipment of very cheap and low ability, and it has very limited calculating or graphics performance and may have the large capacity holder in very limited this locality or not have local large capacity holder.By contrast, each server 402 that is coupled to SAN 403 and multiple RAID 405 can be especially high performance computing system, and in fact, if multiple server with parallel processing configuration cooperation use, there is hardly amount of calculation to bearing and the restriction of graphics capability.In addition, due to low delay video compression 404 and low delay video decompression 412 (by user awareness ground), so the computing capability of server 402 is offered to user.In the time that user presses the button on input equipment 421, the image on display 422 is upgraded (in perception without significant delay) in response to button press, carries out in this locality as game or application software.Therefore, for the computer for unusual low performance or just implement low delay video decompression and the family of the cheap chip of control signal logic 413 or office's client 415, certainly it seems that available remote location is effectively for user provides any computing capability in this locality.This provides the ability of the application program for playing intensive (the being generally new) video-game of the superlative degree, processor and peak performance for user.

Fig. 4 c shows very basic and cheap family or office's client device 465.This equipment is an embodiment according to the family of Fig. 4 a and Fig. 4 b or office's client 415.Its about 2 inches long.It has and the Ethernet jack 462 of Ethernet cable interfaces with Power over Ethernet (PoE), and this equipment obtains its electric power and the connectivity to the Internet thereof from Ethernet jack 462.This equipment can be carried out NAT in the network of network enabled address transition (NAT).In office environment, many new ethernet switch hubs have PoE and directly take PoE to Ethernet jack in office.Under this kind of situation, required is the Ethernet cable from wall type jack to client 465.If connecting, available Ethernet (for example do not transport electric power, there is DSL or cable modem but in family without PoE), exist available cheap wall type " fragment of brick (brick) " (to be also, power supply), it will accept the Ethernet cable of unregulated power and output and have the Ethernet of PoE.

Client 465 contains control signal logic 413 (Fig. 4 a), this bluetooth wireless interface and Bluetooth input equipment 479 interfaces such as keyboard, mouse, game console and/or microphone and/or earphone that is coupled to bluetooth wireless interface.And, an embodiment of client 465 can be with 120fps output video in the situation that being coupled with display device 468, display device 468 can support 120fps video and to a pair of shading eyeglasses 466 signal (conventionally via infrared) then cover another eye alternately to cover an eye for each successive frames.The three-dimensional 3D rendering that the effect that user feels is " jumping out " display screen.Support that this display device 468 of one of this operation is Samsung HL-T5076S.Because be independent for the video flowing of each eye, so in an embodiment who is compressed by host services 210 at two independent video streams, frame is staggered in time, and frame decompresses with two independent decompression process in client 465.

Client 465 also contains low delay video decompression logic 412, the video that its decompression is imported into and audio frequency and export via HDMI (HDMI (High Definition Multimedia Interface)), connector 463, HDMI (HDMI (High Definition Multimedia Interface)), connector 463 are inserted in SDTV (standard definition television) or HDTV (high definition TV) 468, thereby provide video and audio frequency to TV, or be inserted in the monitor 468 of supporting HDMI.If user's monitor 468 is not supported HDMI, can use HDMI to DVI (Digital Visual Interface), but audio frequency will be lost.Under HDMI standard, display capabilities (for example, the resolution of supporting, frame rate) 464 express from display device 468, and then connect 462 via the Internet this information is transferred back to host services 210, therefore host services 210 can make compressed video flow with the form that is suitable for this display device.

Fig. 4 d shows family or office's client device 475, and except client device 475 has more external interfaces, it is identical with family or office's client device 465 shown in Fig. 4 c.And client 475 can be accepted PoE and power, or it can take the external power adapter (not shown) of inserting in wall.Video camera 477 is used client 475USB input that compressed video is provided to client 475, and compressed video uploads to host services 210 for described purposes below by client 475.By utilization below described compress technique low delay compressor reducer is created in camera 477.

Except having the ethernet connector connecting for its Internet, client 475 also has the 802.11g wave point to the Internet.Two kinds of interfaces all can use NAT in the network of supporting NAT.

And except the HDMI connector having for output video and audio frequency, client 475 also has two strands and connects DVI-I connector, two strands connects DVI-I connector and comprises analog output (and have and will provide the VGA standard adapter cable of output).It also has the analog output for composite video and S video.

For audio frequency, client 475 has left/right simulating stereo RCA jack, and for digital audio output, it has TOSLINK (optical fiber) output.

Except to the bluetooth wireless interface of input equipment 479, it also has the USB jack to input equipment for interface.

An embodiment of the inside structure of Fig. 4 e display client 465.All devices or some equipment shown in this figure may be implemented in a programmable logic array, customization ASIC or in several discrete device (Custom Design or ready-made).

The Ethernet 497 with PoE is attached to Ethernet interface 481.The Ethernet 497 that electric power 499 just has PoE obtains and is connected to all the other equipment in client 465.Bus 480 is the same bus of public affairs for the communication between equipment.

(almost any little CPU is suitable to carry out control CPU 483 from the little client controlling application program of flash memory 476, such as the MIPS R4000 series CPU having under the 100MHz of embedded RAM) implement for network (be also, Ethernet interface) protocol stack and all devices in client 465 of also communicating by letter with host services 210, and configure.It is also processed with the interface of input equipment 469 will divide into groups (if desired, together with the customer controller data that are subject to forward error correction protection) and sends back to host services 210.And, control CPU 483 monitor packet communication (for example, grouping be lose or postpone, with and arrive timestamp).This information is sent back to host services 210, with make its consistently monitoring network connect and correspondingly adjust the content of its transmission.Initial during fabrication for flash memory 476 is written into control program for controlling CPU 483 and unique sequence number for particular clients Unit 465.This sequence number allows host services 210 identify customer end Unit 465 uniquely.

Blue tooth interface 484 is wirelessly communicated to input equipment 469 via its antenna (in client 465 inside).

Video decompressor 486 is for being configured to implement the low delay video decompressor of video decompression described herein.Multitude of video decompression apparatus exists, or is off-the-shelf, or as thering is the intellectual property (IP) that can be incorporated into the design in FPGA or customization ASIC.One is provided for the H.264 company of the IP of decoder is the Ocean Logicof Manly of New South Wales,Australia (NSWAustralia).Use the advantage of IP to be: compress technique used herein does not conform to compression standard.Some canonical solution compressor reducers are enough flexibly to be configured to adapt to compress technique herein, but some canonical solution compressor reducers may be really not so.But, the in the situation that of IP, optionally redesigning in decompressor and have full flexibility.

The output of video decompressor is coupled to video output subsystem 487, and video output subsystem 487 is coupled to video the video output terminals of HDMI interface 490.

Audio decompression subsystem 488 or implement with available standard audio decompressor, or it can be embodied as IP, or can for example, in the interior enforcement audio decompression of control processor 483 of can () implementing Vorbis audio decompressor.

Implement the device coupled of audio decompression to audio frequency output subsystem 489, audio frequency output subsystem 489 is the audio output to HDMI interface 490 by audio coupling.

An embodiment of the inside structure of Fig. 4 f display client 475.As visible, (and use if so except additional interface and from the optional outside DC electric power that inserts the power supply adaptor in wall, optional outside DC electric power is replaced the electric power from Ethernet PoE 497) outside, this framework is identical with the framework of client 465.Hereinafter will not repeat and common functional of client 465, but additional functionality is described below.

CPU 483 communicates by letter with extras and configures extras.

WiFi subsystem 482 provides wireless Internet access via its antenna, as substituting Ethernet 497.WiFi subsystem can be purchased from Duo Jia manufacturer, comprises the AtherosCommunications (Aunar He Lusi communication common carrier) of the Santa Clara of California, USA.

USB subsystem 485 provides substituting the Bluetooth communication for wired USB input equipment 479.The suitable standard of USB subsystem and can be easily for FPGA and ASIC, and be often created to and carry out as in the off-the-shelf equipment of other functions of video decompression.

Compare with the video output in client 465, video output subsystem 487 produces the video output of relative broad range.Except HDMI 490 video outputs are provided, it provides DVI-I 491, S-video 492 and synthetic video 493.And, in the time that DVI-I 491 interfaces are used for digital video, display capabilities 464 are transmitted back to and control CPU 483 from display device, to make it can notify to host services 210 ability of display device 478.The total interface being provided by video output subsystem 487 is the interface of suitable standard and easily can uses with many forms.

Audio frequency output subsystem 489 is via digitally output audio and export the audio frequency of analog form via three-dimensional modelling interface 495 of digital interface 494 (S/PDIF and/or Toslink).

Round trip delay Analysis

Certainly,, in order to realize the interests of the last period, user uses the action of input equipment 421 and on display device 420, sees that the round trip time delay between the consequence of this action should be not more than 70 milliseconds-80 milliseconds.This time delay must be considered at the input equipment 421 in customer site 211 to host services 210 and again turn back to customer site 211 to all factors in the path of display device 422.Fig. 4 b illustrates various assemblies and network (signal must be advanced via this assembly or network), and this assembly and network top is timeline, and this timeline is listed expected exemplary time delay in actual enforcement.Note, Fig. 4 b is through simplifying to only show important path route.Other routes of the data of other features that are used for system are below described.Double-headed arrow (for example, arrow 453) instruction round trip time delay and single head arrow (for example, arrow 457) instruction one-way delay, and "～" represents approximate measurement.Should point out, by the real world scenarios that exists listed time delay to reach, but under a large amount of situations, in the U.S., use DSL and cable modem connection to customer site 211, this time delay can be reached in the situation described in next section.And, note, although the cellular radio connectivity to the Internet will be worked really in shown system, most of current American honeycomb fashion data systems (such as, EVDO) cause very high time delay and can not reach time delay shown in Fig. 4 b.But these general principles can be implemented in the following cellular technologies of time delay that may be able to implement this level.

From the input equipment 421 at customer site 211 places, once user activates input equipment 421, (it can be the autonomous device such as Set Top Box just user control signal to be sent to client 415, or it can be software or the hardware in another equipment such as PC or action equipment, carried out), and provide destination-address to arrive host services 210 by its packetizing (in one embodiment taking UDP form) and as dividing into groups.Grouping will also contain the information of control signal from which user that is used to indicate.Then via firewall/router/NAT (network address translation) equipment 443, control signal is forwarded a packet to wan interface 442.Wan interface 442 is for being provided to the interface equipment of customer site 211 by user's ISP (ISP).Wan interface 442 can be cable or dsl data machine, WiMax transceiver, fiber optical transceiver, honeycomb fashion data-interface, power line Internet protocol interface (InternetProtocol-over-powerline interface), or to any other interface in many interfaces of the Internet.In addition, firewall/router/NAT device 443 (and (possibly) wan interface 442) can be incorporated in client 415.This example will be mobile phone, and it comprises the functional software for implementing family or office's client 415, and for for example, via certain standard (, 802.11g) route and be connected to the device of the Internet wirelessly.

Wan interface 442 then routes to control signal " existing a little " 441 of the alleged ISP for user (ISP) herein, and wan interface 442 is for providing the facility of the interface between WAN conveyer and common interconnection network or the private network that is connected to customer site 211.Exist the characteristic of point that the character of the Internet service depending on provided is changed.For DSL, its telephone operator's central office that normally DSLAM is positioned at.For cable modem, it is cable multisystem operator (MSO) front end normally.For cellular system, the control room that it is normally relevant to honeycomb fashion tower.But how no matter there is character a little, its all will be then by control signal Packet routing to common interconnection network 410.Then via by most probable be the interface of fiber optical transceiver interface by control signal Packet routing to wan interface 444 to host services 210.WAN 444 will follow that (it can be implemented by many different modes to logical routing 409 by control signal Packet routing, comprise ethernet switch hub and routing server), the address of logical routing 409 estimating user and control signal is routed to the correct server 402 for given user.

Server 402 is then considered as described control signal the input of the game carried out on server 402 or application software and processes and play or the next frame of application program by described control signal.Once produce next frame, just export video and audio frequency to video compressor 404 from server 402.Can export video and audio frequency to compressor reducer 404 from server 402 via various devices.First, compressor reducer 404 can be created in server 402, therefore can in server 402, implement compression in this locality.Or, can be via to the network (private network between itself or server 402 and video compressor 404, or via the network of the community network such as SAN 403) network connect (such as, Ethernet connects) with the formal output video and/or audio of packetizing.Or, can via video output connector (such as, DVI or VGA connector) video is exported from server 402, and then caught by video compressor 404.And, audio frequency can be output as to digital audio (for example, via TOSLINK or S/PDIF connector) or analogue audio frequency from server 402, analogue audio frequency comes digitlization coding by the audio compression logic in video compressor 404.

Once video compressor 404 has been caught the audio frequency producing from the frame of video of server 402 and during this frame time, video compressor will use below described technique compresses video and audio frequency.Once video and audio frequency are compressed, just by an address by its packetizing to be sent back to user's client 415, and routed to wan interface 444, wan interface 444 is then via common interconnection network 410 route video and audio packet, common interconnection network 410 then routes to video and audio packet the existence point 441 of user's ISP, exist point 441 video and audio packet to be routed to the wan interface 442 at customer site place, video and audio packet are routed to firewall/router/NAT device 443 by wan interface 442, firewall/router/NAT device 443 then routes to client 415 by video and audio packet.

Client 415 decompressed video and audio frequency, and then at the upper display video of display device 422 (or built-in display device of client) and audio frequency is sent to display device 422 or to independent amplifier/loud speaker or to the amplifier/loud speaker being created in client.

For making user feel that described whole process did not lag behind in perception just now, round trip postpones to be less than 70 milliseconds or 80 milliseconds.Some time delays in described round trip path postpone to be subject to host services 210 and/or user's control, and other time delay postpones not to be subject to host services 210 and/or user's control.However, the analysis based on a large amount of real world conditions and test, below measure for approximate.

The one-way transmission time 451 that is used for transmitting control signal is less than 1 millisecond conventionally, conventionally uses the consumer's level firewall/router/NAT interchanger being easy to get on Ethernet and completes in approximately 1 millisecond via the round trip route 452 of customer site.User ISP changes widely its round trip and postpones 453, but the in the situation that of DSL and cable modem supplier, conventionally sees that they are between 10 milliseconds and 25 milliseconds.The business of round trip time delay video capable on common interconnection network 410 is route and whether road exists any fault (and this problem is below being discussed) online and greatly change by how, but common interconnection network provides quite best route and time delay to be determined through the speed (given distance to destination) of optical fiber by light to a great extent conventionally.As further discussed below, determined that 1000 miles as the roughly maximum distance of expecting host services 210 to put away from customer site 211.At 1000 miles (2000 miles of round trips), be approximately 22 milliseconds for signal via the actual transmissions time of the Internet.Be generally the business level optical fiber high-speed interface with insignificant time delay to the wan interface 444 of host services 210.Therefore, common interconnection network time delay 454 is conventionally between 1 millisecond and 10 milliseconds.Can reach being less than in 1 millisecond via unidirectional route 455 time delays of host services 210.Server 402 will calculate in the time being less than a frame time (it is 16.7 milliseconds under 60fps) for playing or the new frame of application program conventionally, and therefore 16 milliseconds be by the rational maximum one-way delay 456 of use.In the best hardware implementation of described video compression and audio compression algorithm, compression 457 can complete in 1 millisecond in this article.In inferior good version, compression can spend and reach 6 milliseconds (certainly, not good enough version can spend the longer time, but described enforcement will affect total time delay of round trip and will need other time delays shorter (for example, can reduce the admissible distance via common interconnection network) to maintain 70 milliseconds of-80 millisecond time-delay targets).Consider the round trip time delay of the Internet 454, user ISP 453 and customer site route 452, therefore remaining is video decompression 458 time delays, video decompression 458 time delays depend on video decompression 458 be performed in specialized hardware, be also performed on client device 415 (such as, PC or action equipment) on software in, the performance of the size of visual display unit and decompression CPU and changing.Conventionally, decompressed for 458 times of cost between 1 millisecond and 8 milliseconds.

Therefore, can determine that by the time delay of being seen all situations worst is in practice added together the user of system shown in Fig. 4 a can expect the round trip time delay of the situation worst of experience.They are: 1+1+25+22+1+16+6+8=80 millisecond.In addition, in fact, (have the explanation that prevents misunderstanding below discussed) in practice, this is roughly the being seen round trip time delay of prototype version (using ready-made Windows PC as client device and the DSL of family and cable modem connection in the U.S.) that uses system shown in Fig. 4 a.Certainly, the situation of situation worst of being better than can cause the time delay of much shorter, but not responsible its developed widely used commerce services.

In order to reach time delay listed in Fig. 4 b via common interconnection network, need video compressor 404 and video decompressor 412 in client 415 (a) to produce and there is the very specifically stream of packets of characteristic from Fig. 4, with make via from host services 210 packet sequence that whole path was produced to display device 422 without undergoing postponing or too much packet loss, and in detail, drop on consistently via user's the Internet and connect (via wan interface 442 and firewall/router/NAT 433) and can be used in user's the constraint of bandwidth.In addition, video compressor must produce enough strong stream of packets, with inevitable packet loss and packet reordering that its tolerable is occurred in normal the Internet and Internet Transmission.

Low delay video compression

In order to complete above-mentioned target, an embodiment adopts new video-frequency compression method, and the method reduces time delay and the peak bandwidth requirement for transmission of video.Before describing this embodiment, will provide the analysis to current video compress technique about Fig. 5 and Fig. 6 a to Fig. 6 b.Certainly,, if user possesses the bandwidth that is enough to process the required data rate of described technology, this technology can be used according to general principle.Note, do not solve audio compression herein, but state audio compression and video compression while and synchronously implement.Meet the prior art audio compression techniques existence for the requirement of this system.

Fig. 5 explanation, for a specific prior art of compressed video, is wherein compressed each indivedual frame of video 501-503 to produce a series of compressed frame 511-513 by compressed logic 520 use specific compression algorithms.An embodiment of this technology is " motion JPEG ", wherein compresses each frame according to joint image expert group (JPEG) compression algorithm based on discrete cosine transform (DCT).Can use various dissimilar compression algorithms, but, this general principle (for example,, taking small echo as basic compression algorithm, such as JPEG-2000) still observed.

A problem of this type compression is: it has reduced the data rate of each frame, but it does not utilize the similarity between successive frame to reduce the data rate that total video flows.For example, as illustrated in fig. 5, suppose the frame rate of 640 × 480 × 24 bits/pixel=640*480*24/8/1024=900 kilobytes/frame (KB/ frame), for the image of given quality, motion JPEG may only compress 1/10 by this stream, thereby produces the data flow of 90KB/ frame.Under 60 frame/seconds, this will need the channel width of 90KB*8 bit * 60 frame/second=42.2Mbps, its for the U.S. now nearly all home interconnect net connect speech and will be high bandwidth, and be too high bandwidth for the Internet connection of many offices.In fact, suppose that it requires constant traffic in the situation that of this kind of high bandwidth, and it will only serve a user, even in office's lan environment, it also when supports the large percentage of bandwidth that consumes 100Mbps ethernet lan the heavy-laden ethernet switch hub of LAN.Therefore, in the time comparing with other compress techniques (all those technology as described below), for the compression inefficiency of sport video.In addition, use the single frame compression algorithm (as JPEG and JPEG-2000) of Lossy Compression Algorithm be created in rest image may the false shadow of unnoticed compression (for example, false shadow in intensive leaf in scene may not be rendered as false shadow, because eye also not exactly knows how intensive leaf presents).But once scene at the volley, false shadow just may be given prominence to, because eye detects the false shadow changing to frame from frame, although false shadow is in the region of scene (in this region, false shadow may be not noticeable in rest image).This causes the perception of " background noise " in frame sequence, and the outward appearance and edge analog TV reception period visible " snowflake " noise that are somebody's turn to do " background noise " are similar.Certainly, the compression of the type still can be used in specific embodiment described herein, but generally speaking, for fear of the background noise in scene, for given perceived quality, needs high data rate (also, low compression ratio rate).

The compression of other types (such as, H.264, or windows media VC9, MPEG2 and MPEG4) all more effective in compressing video frequency flow, because it utilizes the similarity between successive frame.These technology all depend on the identical general technology for compressed video.Therefore,, although will describe H.264 standard, identical General Principle is applicable to various other compression algorithms.H.264 compressor reducer and decompressor can be used in a large number, comprise for compressing x264 Open-source software storehouse H.264 and FFmpeg (a kind of Audio and Video stream scheme) the Open-source software storehouse for decompressing H.264.

Fig. 6 a and Fig. 6 b illustrative examples expressivity prior art compress technique, be wherein compressed into a series of " I frames " 611,671 by compressed logic 620 by the frame of video 501-503 of a series of uncompressed, 559-561; " P frame " 612-613; And " B frame " 670.Vertical axis in Fig. 6 a represents each the gained size (although described frame is drawn not in scale) in encoded frame substantially.As mentioned above, those skilled in the art understands well and uses the video of I frame, B frame and P frame to write code.In brief, the compression based on DCT (being similar to compressed jpeg image as above) that I frame 611 is complete unpressed frame 501.The size of P frame 612-613 is significantly less than the size of I frame 611 conventionally, because it utilizes the data in previous I frame or P frame; Also, its data that contain the change between the previous I frame of instruction or P frame.Except B frame uses frame in reference frame subsequently and (possibly) before the frame in reference frame, B frame 670 is similar to P frame.

For following discussion, will suppose that desired frame rate was 60 frame/seconds, each I frame is about 160Kb, mean P frame and B frame are 16Kb and produced a new I frame every one second.Under this group parameter, mean data rate will be: 160Kb+16Kb*59=1.1Mbps.This data rate suitably drops in the maximum data rate for connecting to the many current broadband internet of family and office.This technology is also tended to avoid from the background noise problem of intraframe coding only, because P frame and B frame are followed the trail of the difference between frame, therefore compresses false shadow and tends to do not present and disappear to frame from frame, thereby reduce background noise problem as described above.

A problem of the compression of the above-mentioned type is: although for example, mean data rate relatively low (, 1.1Mbps), single I frame may spend several frame times and transmit.For example, use prior art, 2.2Mbps network connects (for example, DSL or cable modem, it has the 2.2Mbps peak value from the maximum usable data rates 302 of Fig. 3 a) conventionally will enough make video flow with 1.1Mbps, the I frame of a 160Kbps of every 60 frames.This will complete by making decompressor, before decompressed video, 1 second video be entered to queue.In 1 second, data that will transmission 1.1Mb, it will easily adapt to by 2.2Mbps maximum usable data rates, 50% also like this even if supposition available data rate may periodically decline nearly.Unfortunately, this prior art approach will cause time delay in 1 second of video due to 1 second video buffer at receiver place.This kind of delay is for example, for many prior art application programs (, the playback of linear video) enough, but is greater than the fast motion video-game of time delay of 70 milliseconds-80 milliseconds for not tolerable and time delay that Yan Qiwei extremely grows.

Eliminate 1 second video buffer if attempt, it will not can cause reducing for enough time delays of fast motion video-game.For example, as discussed previously, the use of B frame need to receive front all B frames and the I frame of I frame.If 59 non-I frames are roughly divided in supposition between P frame and B frame, will there are at least 29 B frames and can show the I frame receiving before any B frame.Therefore, no matter the available bandwidth of channel how, all needs the delay of 29+1=30 each 1/60 second duration of frame, or the time delay of 500 milliseconds.Obviously, this time compole long.

Therefore, other method will be eliminate B frame and only use I frame and P frame.(one of them consequence is: for given quality level, data rate will increase, but for the purpose of the consistency in this example, the size that continues the each I frame of supposition is that the size of 160Kb and mean P frame is 16Kb, and therefore data rate is still 1.1Mbps).The method has been eliminated the inevitable time delay of being introduced by B frame, because the decoding of each P frame only depends on the frame previously receiving.The problem that the method still exists is: I frame is more much bigger than mean P frame, so that in low bandwidth channel (as typical in most of families and in many offices), essence time delay is added in the transmission of I frame.This is illustrated in Fig. 6 b.Video stream data speed 624 is lower than available maximum data speed 621 (except for I frame), and wherein the required peak data rate 623 of I frame far exceedes available maximum data speed 622 (and even exceeding specified maximum data rate 621).The required data rate of P frame is less than available maximum data speed.Even if the available maximum data speed peak value at 2.2Mbps stably remains on its 2.2Mbps peak rate, also will spend 160Kb/2.2Mb=71 millisecond and transmit I frame, if and available maximum data speed 622 50% (1.1Mbps) that decline, will spend 142 milliseconds and transmit I frame.Therefore, the time delay in transmission I frame is by the somewhere of dropping between 71 milliseconds and 142 milliseconds.The time delay (described time delay is adding up 70 milliseconds under situation worst) of identifying in Fig. 4 b is added in this time delay to, therefore, this activates the moment of input equipment 421 until image is presented in the total round trip time delay for 141-222 millisecond display device 422 by causing from user, and it is high.And if available maximum data speed drops to lower than 2.2Mbps, time delay will further increase.

The peak data rate 623 that is also noted that far to exceed available data rate 622 makes ISP " obstruction " conventionally have serious consequence.Equipment in different I SP will differently show, but in the time receiving grouping with the data rate more much higher than available data rate 622, following behavior is quite general in the middle of DSL and cable modem ISP: (a) make packetization delay (introducing time delay) by grouping being entered to queue, (b) abandon some or all of groupings, (c) will connect stop using a time cycle (most probable, because ISP worries it for malicious attack, is attacked such as " denying service ").Therefore be not, as feasible option taking full data rate transport packet stream (thering is the characteristic such as described characteristic shown in Fig. 6 b).Can at host services 210 places, peak value 623 be entered to queue and to send lower than the data rate of available maximum data speed, thereby introduce the unacceptable time delay described in the last period.

In addition, in Fig. 6 b, shown video stream data speed sequence 624 is unusual " tame (tame) " video stream data speed sequence, and will be this kind of data rate sequence of expecting generation from the video of video sequence due to compression, this video sequence does not change very large and (for example has considerably less motion, as will be general in video conference call, in video conference call, camera is in a fixed position and has a considerably less motion, and for example, object in scene (people who, takes one's seat talks) shows less physical activities).

In Fig. 6 c shown video stream data speed sequence 634 for the video from thering is much more action (such as, may in film or video-game or in certain application software, produce) expect visible type sequence.Note, except I frame peak 633, also exist quite greatly and exceed in many occasions available maximum data speed P frame peak (such as, 635 and 636).Although described P frame peak is also be not as generally quite large as I frame peak, it still greatly consequently can not be transported with full data rate by channel, and as I frame peak, P frame peak must transmit (increasing whereby time delay) lentamente.

For example, in high-bandwidth channels (, 100Mbps LAN, or the connection of high bandwidth 100Mbps private) upper, network can be tolerated the large peak value such as I frame peak 633 or P frame peak 636, but in principle, can maintain low delay.But, described network (for example shares in the middle of many users that are everlasting, in office environment), and should " having peak " data the performance of LAN will be affected, especially for example share, in the situation that connects (, from remote data center to office) through routing to private in network service.First, remember that this example is the example of the video flowing of the relative low resolution of 640 × 480 pixels under 60fps.The HDTV fluid capacitance of under 60fps 1920 × 1080 is easily processed by modern computer and display, and 2560 × 1440 resolution displays under 60fps day by day can for example, with (, 30 of Apple company " display).Use H.264 compression, the high action video sequence of 1920 × 1080 under 60fps may need 4.5Mbps to obtain reasonable quality level.If supposition I frame peak is 10 times of nominal data speed, it will produce 45Mbps peak value, and less but still sizable P frame peak.For example, if several users just at same 100Mbps network (, private network between office and data center is connected) upper receiver, video stream, easily see from the peak value of some users' video flowing and can how to aim at by chance, thereby flood the bandwidth of network, and may flood the bandwidth of supporting the base plate of user's interchanger on network.Even under the situation of ultrahigh speed Ethernet, if enough users have enough peak values of aiming at simultaneously, it can flood network or network switch.In addition,, once 2560 × 1440 resolution videos become more common, average video stream data rate may be just 9.5Mbps, thereby perhaps produces 95Mbps peak data rate.Much less, the 100Mbps between data center and office is connected (its now for especially fast connection) and will be routed by the peak communication from sole user completely.Therefore,, even if LAN and private network connect there being the mobile video in peak can have higher tolerance, the mobile video with peak value is also unwanted and may needs special project and the adaptation of the IT department of office.

Certainly, for normal linearity video application, this problem is not problem, because data rate transfer point " through smoothing " and for the data of each frame lower than maximum usable data rates 622, and before decompression I frame, P frame and B frame sequence, the buffer in client stores I frame, P frame and B frame sequence.Therefore, the data rate on network keeps the mean data rate close to video flowing.Regret, this introduces time delay, even if do not use B frame, for such as needing for the video-game of fast-response time and the low delay application program of application program, this is also unacceptable.

Usually to use the technology that " constant bit-rate " (CBR) encodes that is known as for a prior art solution that alleviates the video flowing with peak value.Although term CBR appears to imply that all frames are compressed to have same bits speed (is also, size), but its often mention for compression example, compression example in, allow the maximum bit rate of the frame (under our situation, 1 frame) of crossing over given number.For example, under the situation of Fig. 6 c, for example, if coding is applied to CBR constraint (its bit rate is limited to () specified maximum data rate 621 70%), the compression of compression algorithm each in described frame by restriction, with make conventionally by with specified maximum data rate 621 70% with on any frame of compressing will compress with less bit.This result is: the picture quality that the frame that makes conventionally more needs bits to be maintained given quality level " is extremely lacked " to bit and described frame is by than not needing than the picture quality grain of other frames of more than 70% bit of specified maximum data rate 621.The method for the compressed video of particular type (wherein (a) expection less physical activities or scene change and (b) user can accept periodic degrading quality) can produce acceptable result.The good example that is applicable to the application of CBR is video conference call, because there is less peak value, and the in the situation that of the temporary transient degradation of quality (for example, if make camera pan, thereby cause remarkable scene motion and large peak value, during sweeping, may not exist enough bits to compress for high quality graphic, it will cause the picture quality of degradation), most of users can accept.Regret, CBR is not applicable to well having the scene of high complexity or moves in a large number and/or rationally many other application of constant quality level of needs.

The low delay compressed logic 404 that used in one embodiment uses some different technologies to solve the mobile compressed video of low delay and maintains high-quality many problems simultaneously.First, low delay compressed logic 404 only produces I frame and P frame, the needs of each B frame thereby several frame times of alleviation wait are decoded.In addition, as illustrated in Fig. 7 a, in one embodiment, low delay compressed logic 404 is divided into the frame 701-760 of each uncompressed more a series of " image block (tile) " and each image block is individually encoded to I frame or P frame.In this article this group of compressed I frames and P frame are called to " R frame " 711-770.In Fig. 7 a, in shown particular instance, the frame of each uncompressed is divided into again to 16 image blocks of 4 × 4 matrixes.But described general principle is not limited to any specific point mechanism again.

In one embodiment, frame of video is divided into many image blocks by low delay compressed logic 404, and by an image block coding from each frame (be also, compression) for I frame (be also, by this image block compression, as its be full images big or small independent frame of video of 1/16, and for the boil down to I frame compression of this " mini " frame) and residual image piece is encoded to P frame (also, compressing for the boil down to P frame of each " mini " 1/16 frame).Compressed image block for I frame and compressed image block for P frame will be known as respectively " I image block " and " P image block ".Change the image block for I image block to be encoded along with each subsequent video frame.Therefore, in given frame time, in the described image block in frame of video, only an image block is I image block, and the surplus P image block that the rest is in described image block.For example, in Fig. 7 a, the image block 0 of the frame 701 of uncompressed is encoded is that I image block I0 and remaining 1-15 image block are encoded for P image block (P1 to P15) is to produce R frame 711.In the frame of video 702 of next uncompressed, the image block 1 of the frame 701 of uncompressed is encoded is that I image block I1 and remaining image block 0 and 2 to 15 are encoded for P image block (P0, and P2 to P15) is to produce R frame 712.Therefore, little by little staggered in time in successive frames for I image block and the P image block of image block.This process continues, until till producing R image block 770 (in matrix most end image block encoded be I image block (also, I15)).This process then restarts, thereby produces another R frame such as frame 711 (also, for image block 0, coding I image block) etc.Although undeclared in Fig. 7 a, in one embodiment, a R frame of the video sequence of R frame only contains I image block (also, to make P frame subsequently have reference image data (starting to calculate motion from it)).Or in one embodiment, initiating sequence uses the I image block type sample same with normal phase, but do not comprise the P image block of the described image block for not yet encoding together with I image block.In other words, before an I image block arrives not together with any data specific image block of encoding, thereby avoid the startup peak value in the video stream data speed 934 in Fig. 9 a, it is below further describing.In addition, as described below, various different sizes and shape can be used for described image block and still observe described general principle simultaneously.

Video decompression logic 412 each image block that decompresses of carrying out in client 415, be the independent video sequence of little I frame and P frame as it, and then each image block is reproduced to the frame buffer to driving display device 422.For example, use to carry out decompress(ion) from the I0 of R frame 711 to 770 and P0 and contract and reproduce the image block 0 of video image.Similarly, use from I1 and the P1 of R frame 711 to 770 and rebuild image block 1, etc.As mentioned above, the decompression of I frame and P frame is many known in this technology, and the decompression of I image block and P image block can have been carried out by the multiple execution individualities that make video decompressor in client 415.Although multiplicative process appears to increase the computation burden in client 415, but in fact it can not increase the computation burden in client 415, because image block pro rata less (for the number of extra process) itself, therefore the number of shown pixel is identical, as the I frame and the P frame that exist one to process and use traditional full size.

This R frame technique significantly alleviates the bandwidth peak (illustrated in Fig. 6 b and Fig. 6 c) being conventionally associated with I frame, and giving framing because any is to be mainly made up of the common P frame less than I frame.For example, again suppose that typical I frame is 160Kb, in Fig. 7 a, the I image block of each in illustrated frame will be roughly 1/16 or 10Kb of this amount.Similarly, suppose that typical P frame is 16Kb, can be roughly 1Kb for each the P frame in the illustrated image block of Fig. 7 a.Final result is the R frame of about 10Kb+15*1Kb=25Kb.Therefore, each 60 frame sequence will be 25Kb*60=1.5Mbps.Therefore, under 60 frame/seconds, this can maintain needs the channel of the bandwidth of 1.5Mbps, but because I image block is to run through 60 frame periods to disperse to make to have much lower peak value.

Note, in previous example, under the identical tentation data rate behavior for I frame and P frame, mean data rate is 1.1Mbps.This is because in previous example, only introduce new I frame every 60 frame times, and in this example, 16 image blocks that form I frame circulate in 16 frame times, and therefore, introduce the equipollent of I frame every 16 frame times, thereby cause slightly high mean data rate.However, but in practice, introduce I frame more frequently and can't increase linearly data rate.This is due to the fact that P frame (or P image block) main code is from the extremely difference of next frame of previous frame.Therefore, if previous frame and next frame are quite similar, P frame will be very little, if previous frame is quite different from next frame, P frame will be very large.But because P frame is to derive from previous frame to a great extent, but not derive from actual frame, so the encoded frame of gained for example may contain, than the many mistakes (, visual artifacts) of I frame of bit with enough numbers.In addition,, in the time that a P frame is followed another P frame, can there is mistake cumulative (in the time there is long P frame sequence, becoming even worse).Now, most advanced and sophisticated video compressor will detect the quality of image in the fact of the rear degradation of a sequence P frame, and if desired, and it distributes to P frame subsequently to improve quality by more bits, if or its be the most effective course of action, with I frame replacement P frame.Therefore, for example, when using long P frame sequence (, 59 P frames, as above previously in example) time, particular words there is large amount of complex degree and/or when motion when scene, conventionally, need to more bits (because it becomes farther apart from I frame) for P frame.

Or from relatively watching and a little see P frame, the P frame of closely following I frame tends to need to be than apart from the farther few bit of P frame of I frame.Therefore, in Fig. 7 a, in shown example, do not have P frame to separate 15 frames farther (before I frame) than apart from I frame, and in previous example, P frame may separate 59 frames from I frame.Therefore,, under I frame condition more frequently, P frame is less.Certainly, definite relative size changes the character based on video flowing, but in the example of Fig. 7 a, if I image block is 10Kb, the size of P image block on average can be only 0.75kb, thereby causes 10Kb+15*0.75Kb=21.25Kb, or under 60 frame/seconds, data rate will be 21.25Kb*60=1.3Mbps, or than thering is the I frame and then data rate of the stream of 59 P frames high approximately 16% subsequently under 1.1Mbps.Again, will change depending on video sequence for the relative result between described two kinds of methods of video compression, but conventionally, we find by rule of thumb, for given quality level, use R frame to need how about 20% bit than use I/P frame sequence.But certainly, R frame reduces peak value sharp, this makes video sequence available under the time delay much smaller than I/P frame sequence.

The character of visual frequency sequence, the reliability of channel and available data rate and configure R frame with multitude of different ways.In alternate embodiment, in 4 × 4 configurations, use the image block of the number that is different from 16.For example, can in 2 × 1 or 1 × 2 configuration, use 2 image blocks, can in 2 × 2,4 × 1 or 1 × 4 configuration, use 4 image blocks, can in 3 × 2,2 × 3,6 × 1 or 1 × 6 configuration, use 6 image blocks or can in 4 × 2 (as shown in as Fig. 7 b), 2 × 4,8 × 1 or 1 × 8 configuration, use 8 image blocks.Note, image block does not need for square, and frame of video needn't be also square, or rectangle even.Image block can be resolved into best and to be applicable to the video flowing that uses and the whatsoever shape of application program.

In another embodiment, the circulation of I image block and P image block is not locked to the number of image block.For example, in 8 image blocks 4 × 2 configurations, still can be as illustrated in Fig. 7 b and use 16 cyclic sequences.The frame 721,722,723 of the uncompressed of order is respectively hung oneself and is divided into 8 image block 0-7, and each image block is compressed individually.From R frame 731, only image block 0 is compressed to I image block, and residual image piece is compressed to P image block.For R frame 732 subsequently, all 8 image blocks are compressed to P image block, and then for R frame 733 subsequently, image block 1 is compressed to I image block and other image blocks are all compressed to P image block.In addition, so proceed sequence for 16 frames, only produce an I image block every a frame, therefore during the 15th frame time (not shown in Fig. 7 b) and during the 16th frame time, produce for the most end I image block of image block 7 (using all P image block compression R frames 780).Then, sequence is again compressed to I image block and other image blocks with image block 0 and is compressed to P image block and starts.As in the previous embodiment, the first frame of whole video sequence will be I image block conventionally, to be provided for from the reference of this some P image block forward.The circulation of I image block and P image block does not even need the even multiple for the number of image block.For example, the in the situation that of 8 image blocks, using before another I image block, having that to can be all after each frame of an I image block be all 2 frames of P image block.In another embodiment, (for example), if the specific region of known screen has more do more physical exercises (needing I image block more frequently), for example, and other regions are more static (, show the mark of game) (needing I image block more frequently),, compared with other image blocks, can more frequently specific image block be sorted together with I image block.In addition, although illustrate that in Fig. 7 a-Fig. 7 b each frame has single I image block, the multiple I image blocks (depending on the bandwidth of transmission channel) of can encoding in single frame.On the contrary, particular frame or frame sequence can transmit in the situation without I image block (also, only P image block).

The reason that the method for the last period suitably works is: appear to cause larger peak value although do not have the I image block of crossing over each single frame and disperse, the behavior of system is so not simple.Because each image block is to separate and compress with other image blocks, so in the time that image block diminishes, it is more ineffective that the coding of each image block can become, because the compressor reducer of given image block can not utilize similar characteristics of image and similar motion from other image blocks.Therefore, compare with screen divider being become to 8 image blocks, become 16 image blocks conventionally will cause more not efficient coding screen divider.But if become 8 image blocks and its to cause the data of introducing a complete I frame every 8 frames (but not every 16 frames) screen divider, it causes much higher generally data rate.Therefore,, by introducing a complete I frame every 16 frames (but not every 8 frames), reduced total data rate.And, by using 8 larger image blocks (but not 16 less image blocks), reducing total data rate, it also alleviates the data peaks being caused by larger image block to a certain degree.

In another embodiment, low delay video compression logic 404 known features by the video sequence based on to be compressed in Fig. 7 a and Fig. 7 b and by setting the ongoing analysis of pre-configured or picture quality based in each image block and automatically control to the distribution of the bit of the each image block in R frame.For example, in some contest video-games, the front of player's automobile (it is for relatively nonmotile in scene) occupies the major part of the Lower Half of screen, and the first half of screen is filled just approaching road, building and landscape completely, and it almost always at the volley.If the bit of equal number is distributed to each image block by compressed logic 404, the image block (image block 4-7) in the Lower Half of the screen in the frame 721 of the uncompressed in Fig. 7 b compresses (image block 0-3) the high quality of the image block in the first half of the screen in the frame of the uncompressed with than in Fig. 7 b 721 conventionally.If this special scenes of known this particular game or game has described characteristic, the configurable compressed logic 404 of the operator of host services 210 is to distribute to more bits the image block (compared with distributing to the bit of image block at bottom place of screen) at the top of screen.Or, compressed logic 404 can be after the condensed frame compression quality of estimated image piece (use one or more in much compression qualities tolerance, such as peak signal noise than (PSNR)), if and determine on special time window, specific image block produces better quality result consistently, it is little by little distributed to more bits and produces compared with the image block of low quality result, until various image block reaches the quality of similar level.In alternate embodiment, compressor logic 404 allocation bit to reach better quality in specific image block or image block group.For example, it can provide preferably general perceives outward appearance, to have the quality higher than edge at the center of screen.

In one embodiment, in order to improve the resolution of specific region of video flowing, what the less image block of video compression logic 404 use carried out encoded video streams has relatively many scene complexities and/or a region (region of few scene complexity relative to having of video flowing and/or motion is compared) of motion.For example, as illustrated in fig. 8, the less image block of use around the mobile personage 805 in a region of a R frame 811 (may and then there are subsequently a series of R frames (not shown) of identical image block size).Then, in the time that personage 805 moves to the new region of image, the less image block of use around this new region in another R frame 812, as described.As mentioned above, various different sizes and shape can be used as " image block " and still observe described general principle simultaneously.

Although circulation I/P image block as described above reduces in fact the peak value in the data rate of video flowing, but it also not exclusively eliminates peak value, especially change fast or the situation of the video image (such as occurring under film, video-game and a certain application software) of high complexity under.For example, during unexpected scene transitions, complex frames may be followed diverse another complex frames subsequently.Even if several I image blocks can be led over only several frame times of scene transitions, it does not also help in this case, because the material of new frame was with previously I image block was irrelevant.In this case (even and not all change, under other situations that great amount of images also changes), video compressor 404 will determine many (if not all) P image block will more effectively be write to code for I image block, and what cause is the very large peak value in the data rate of described frame.

As discussed previously, it is only the situation that connects (and many offices connect) for most consumers level the Internet, and it only can not " stop up " to exceed and in Fig. 6 c, be shown as 622 available maximum data speed and the data of specified maximum data rate 621.Note, specified maximum data rate 621 (for example, " 6Mbps DSL ") is essentially for considering and buys user's the sales figure that the Internet connects, but its guaranteed performance level not conventionally.For the object of this application, it is incoherent, because we only pay close attention to the available maximum data speed 622 while video being flowed via connection.Therefore,, in Fig. 9 a and Fig. 9 c, in the time describing the solution of spike problem, omit specified maximum data rate from curve chart, and only show available maximum data speed 922.Video stream data speed must not exceed available maximum data speed 922.

For head it off, the first thing that video compressor 404 carries out is to determine peak data rate 941, the data rate that it can stably be processed for channel.This speed can be determined by many technology.A kind of this technology is that the data rate test stream all the more uprising is sent to client 415 (in Fig. 4 a and Fig. 4 b) gradually from host services 210, and client is provided the feedback of the level about packet loss and time delay to host services.In the time that packet loss and/or time delay start to show sharp-pointed increase, it is the instruction that reaches available maximum data speed 922.Afterwards, host services 210 can little by little reduce the data rate of test stream, has received and has tested stream (acceptable level of packet loss and time delay approaches minimum) until client 415 was reported in the rational time cycle.This determines peak value maximum data rate 941, and it is then by the peak data rate with acting on the video that flows.As time goes on, peak data rate 941 (for example will fluctuate, if another user in family starts seriously to use the Internet to connect), and client 415 need to monitor peak data rate 941 is to watch packet loss or time delay whether to increase (instruction available maximum data speed 922 drops to lower than previous determined peak data rate 941) consistently, and peak data rate 941 if so.Similarly, if As time goes on, client 415 finds that packet loss and time delay remain on optimum level, it can ask video compressor for example to increase lentamente data rate, to watch available maximum data speed (whether to increase, if another user in family has stopped the serious use that the Internet is connected), and again wait for until packet loss and/or higher time delay instruction have exceeded available maximum data speed 922, and can again find the reduced levels for peak data rate 941, but the level before the data rate that this reduced levels may increase higher than test.Therefore, can be by using this technology (and similar its other technologies) to find peak data rate 941, and optionally and periodically adjust.Peak data rate 941 is determined and can be used so that video flows to user's maximum data rate by video compressor 404.Can and/or be implemented in host services 210 at customer site 211 places for the logic of determining peak data rate.At customer site 211 places, client device 415 is carried out and is calculated to determine peak data rate and this communication is back to host services 210; At host services 210 places, the server 402 at host services place is carried out to calculate and for example, is determined peak data rate with the statistics based on being received from client 415 (, peak value loss, time delay, maximum data rate etc.).

Fig. 9 a shows to have the instance video stream data rate 934 of essence scene complexity and/or motion, and it is the circulation I/P image block compress technique that uses previously to have described and be illustrated in Fig. 7 a, Fig. 7 b and Fig. 8 and producing.Video compressor 404 is configured and to export compressed video lower than the mean data rate of peak data rate 941, and notes, the most of the time, video stream data speed keeps below peak data rate 941.The circulation I/P image block compression that relatively shows of shown video stream data speed 634 in data rate 934 and Fig. 6 c (its be use I/P/B or I/P frame and produce) produces much level and smooth data rate.But under 2 times of peak values 952 of frame (it approaches 2 times of peak data rate 942) and 4 times of peak values 954 of frame (it approaches 4 times of peak data rate 944), data rate still exceedes peak data rate 941, and this is unacceptable.In practice, even for the high action video of the video-game from quick change, the peak value that exceedes peak data rate 941 also occurs in the frame that is less than 2%, and the peak value that exceedes 2 times of peak data rate 942 seldom occurs, and the peak value that exceedes 3 times of peak data rate 943 seldom occurs.For example, but in the time that it occurs really (, between scene tour), its required data rate must produce the video image of good quality.

A mode of head it off is to configure simply video compressor 404 to make its maximum data rate be output as peak data rate 941.Unfortunately, the gained video output quality of peak value image duration is bad, because compression algorithm " extremely lacks " bit.What cause compresses false shadow for occurring in the time there is sudden transition or rapid movement, and in time, user starts to recognize: in the time there is flip-flop or rapid movement, vacation shadow always occurs suddenly, and it can become quite disagreeable.

Although people's vision system is quite responsive to the visual artifacts occurring during flip-flop or rapid movement, under described situation, detect frame rate to reduce not be very responsive.In fact, in the time that described flip-flop occurs, appear people's vision system and be absorbed in and follow the trail of described change, and if frame rate temporarily drop to 30fps and then turn back to immediately 60fps from 60fps, people's vision system can not be noticed.In addition,, under the situation of transformation (as unexpected scene change) very sharply, if frame rate drops to 20fps or 15fps and then turn back to immediately 60fps even, people's vision system can not be noticed.As long as frame rate reduces only to occur once in a while, for people observer, appear video constantly to carry out with 60fps.

Utilize this characteristic of people's vision system by technology illustrated in Fig. 9 b.Server 402 (b) produces the video output streams of uncompressed from Fig. 4 a and Fig. 4 with stabilizer frame speed (in one embodiment, under 60fps).Timeline shows each frame 961-970 output in each 1/60 second.From frame 961, export the frame of video of each uncompressed to low delay video compressor 404, low delay video compressor 404 compresses described frame in the time that is less than a frame time, produces the compressed frame 1981 for the first frame.Larger or less through producing for data visualization many factors as described earlier of compressed frame 1981.If data are enough little thus can peak data rate 941 at a frame time (1/60 second) or be less than and transmit it to client 415 in a frame time, during the transmission time (xmit time) 991 (length of the arrow of the duration in instruction transmission time) by its transmission.In next frame time, server 402 produces the frame 2962 of uncompressed, is compressed into compressed frame 2982, and transmits it to client 415 with peak data rate 941 during being less than the transmission time 992 of a frame time.

Then,, in next frame time, server 402 produces the frame 3963 of uncompressed.In the time compressing the frame 3963 of uncompressed by video compressor 404, the compressed frame of gained 3983 is the many data of data than can peak data rate 941 transmitting in a frame time.Therefore, during the transmission time (2 times of peak values) 993, transmitted, it occupies a part for all frame times and next frame time.Now, during next frame time, server 402 produces the frame 4964 of another uncompressed and is exported to video compressor 404, but data are left in the basket and illustrate by 974.This is because video compressor 404 is configured to ignore the frame of video of other uncompressed that arrive in the time that it still transmits previously compressed frame.Certainly, the video decompressor of client 415 will fail to receive frame 4, but it continues simply display frame 3 on display device 422 and lasts 2 frame times (also, temporarily frame rate being decreased to 30fps from 60fps).

For next frame 5, server 402 is exported the frame 5965 of uncompressed, be compressed into compressed frame 5985 and during the transmission time 995 in 1 frame by its transmission.The video decompressor decompressed frame 5 of client 415 is also shown on display device 422.Then, server 402 is exported the frame 6966 of uncompressed, and video compressor 404 is compressed into compressed frame 6986, but now the data of gained are very large.During the transmission time (4 times of peak values) 996, transmit compressed frame with peak data rate 941, carry out transmission frame but spend almost 4 frame times.During ensuing 3 frame times, video compressor 404 is ignored 3 frames from server 402, and the decompressor of client 415 stably remains on frame 6 and on display device 422, lasts 4 frame times (also, temporarily frame rate being decreased to 15fps from 60fps).Then last, server 402 output frames 10970, video compressor 404 is compressed into compressed frame 10987, and during the transmission time 997, transmitted, and the decompressor decompressed frame 10 of client 415 and be shown on display device 422 and again video restart with 60fps.

Note, although video compressor 404 has abandoned the frame of video of the video flowing that free server 402 produces, but it can not abandon voice data (no matter audio frequency is with what form to come), and video compressor 404 continues audio compressed data and transmits it to client 415 in the time abandoning frame of video, client 415 continues decompression voice data and audio frequency is provided to the whatsoever equipment being used by user with plays back audio.Therefore during abandoning the cycle of frame, audio frequency continues and does not weaken.Compared with compressed video, the bandwidth of the compressed relatively little percentage of audio frequency consumption, and therefore can not have considerable influence to total data rate.Although all undeclared in any one in data rate figure, in peak data rate 941, always there is the data rate capacity through being preserved for compressed audio stream.

Select just to illustrate at the example described in Fig. 9 b how frame rate declines between the data rate peak period, but unaccounted is in the time using previous described circulation I/P image block technology, described data rate peak value and with the frame abandoning of generation little, even high scene complexity/high action sequence (such as in video-game, film and certain application software, occur those) during also like this.Therefore, the frame rate reducing is rare and temporary transient, and people's vision system can not detect them.

Be applied to video stream data speed illustrated in Fig. 9 a if described frame rate is just reduced to mechanism, the video stream data speed of gained is described in Fig. 9 c.In this example, 2 times of peak values 952 have been decreased to 2 times of peak values 953 of planarization, and 4 times of peak values 955 have been decreased to 4 times of peak values 955 of planarization, and whole video stream data speed 934 remains in or lower than peak data rate 941.

Therefore, use technology as described above, can connect and transmit high action video stream with low delay via common interconnection network and consumer's level the Internet.In addition, at LAN (for example, 100Mbs Ethernet or 802.11g wireless network) upper or private network is (for example, 100Mbps between data center and office is connected) on office environment in, can flow without transmitting high action video under peak condition, for example, (to make multiple users, with 1920 × 1080 under 4.5Mbps transmission 60fps) can use LAN or shared private data to connect, and the peak value that do not make to overlap floods (overwhelm) network or network switch base plate.

Data rate is adjusted

In one embodiment, the available maximum data speed 622 of host services 210 initial evaluation channels and time delay are to be identified for the proper data speed of video flowing and then dynamically to adjust data rate in response to this.In order to adjust data rate, host services 210 can (for example) be revised image resolution ratio and/or the number of pictures per second of the video flowing to client 415 to be sent.And, the quality level of the compressed video of host services capable of regulating.For example, in the time changing the resolution of video flowing (, from 1280 × 720 resolution to 640 × 360), the video decompression logic 412 in client 415 can increase on display screen, to maintain identical image size in proportion by image.

In one embodiment, under the situation exiting completely at channel, host services 210 is by suspension of game play.Under the situation of multiplayer, host services to this user of other user report exited game and/or by suspension of game play for other users.

The grouping that abandons or postpone

In one embodiment, if data are because the packet loss between video compressor 404 and client 415 in Fig. 4 a or Fig. 4 b is lost, or owing to arriving evening so that can not decompress and meet through the grouping of the delay requirement of decompressed frame and be lost by out of order reception, video decompression logic 412 can alleviate visual artifacts.In mobile I/P frame is implemented, if there is the grouping of losing/postponing, whole screen is influenced, thereby may cause screen fully charge cycle a period of time or show the wide visual artifacts of other screens.For example, if the grouping of losing/postponing causes the loss of I frame,, before receiving new I frame, decompressor will lack the reference of all P frames for following.If lose P frame, its P frame for whole screen that impact is followed.Before occurring depending on I frame, how long have, this will have long or shorter visual impact.Use as in Fig. 7 a and Fig. 7 b as shown in staggered I/P image block, the grouping of losing/postponing unlikely affects whole screen, because it only affects image block contained in affected grouping.If the data of each image block are to send in individual packets, if packet loss, it only affects an image block.Certainly, the duration of visual artifacts will depend on whether the grouping of I image block loses and before I image block occurs, will spend how many frames in the situation that P image block is lost.But, suppose different images piece on screen be by I frame very continually (may each frame) upgrade, even if the image block on screen is influenced, other image blocks also may be unaffected.In addition, for example, if a certain event causes some groupings and (loses simultaneously, in abutting connection with the spiking of the temporary transient interruption data flow in the electric power of DSL line), some image blocks will be subject to larger impact than other image blocks, but because some image blocks promptly upgrade the I image block by new, so it is only temporarily influenced.And, in the situation that mobile I/P frame is implemented, not only I frame is most critical frame, and I frame is very big, if therefore there is the event of the grouping that abandons/postpone of causing,, compared with much smaller I image block, the higher probability of the influenced existence of I frame (is also, if any part of I frame is lost, I frame may be able to not decompress).Due to all described reasons, compared with the situation of I/P frame, in the time that grouping is dropped/postpones, use I/P image block to cause much smaller visual artifacts.

An embodiment attempts the effect that reduces lost packets in TCP (transmission control protocol) grouping or UDP (User Datagram Protoco (UDP)) grouping by compressed image block is packaged in intelligently.For example, in one embodiment, as possible, aim at boundaries of packets by image block.How Figure 10 a explanation can be packaged in image block in a series of grouping 1001-1005 in the situation that not implementing this feature.In specific words, in Figure 10 a, image block is crossed boundaries of packets and through encapsulation so that the loss of single packet cause the loss of multiple frames inefficiently.For example, if grouping 1003 or 1004 is lost, lose three image blocks, cause visual artifacts.

By contrast, Figure 10 b explanation is for being packaged in image block intelligently the image block encapsulation logical one 010 of the interior effect with minimizing packet loss of grouping.First, image block encapsulation logical one 010 is aimed at image block with boundaries of packets.Therefore, image block T1, T3, T4, T7 and T2 respectively with grouping 1001-1005 boundary alignment.Image block encapsulation logic also attempts in possible more effective mode, image block to be combined in grouping, and does not cross boundaries of packets.Based on each the size in image block, image block T1 and T6 are combined in a grouping 1001; T3 and T5 are combined in a grouping 1002; Image block T4 and T8 are combined in a grouping 1003; Image block T8 is added into grouping 1004; And image block T2 is added into grouping 1005.Therefore,, under this scheme, single packet loss will cause the loss of no more than 2 image blocks (but not as 3 image blocks illustrated in Figure 10 a).

In Figure 10 b, an additional benefit of shown embodiment is: image block is to transmit with its different order shown in image.If contiguous grouping is because the same event (it will affect the region of keeping off each other on screen) of disturbing transmission is lost, this mode produces more unnoticed false shadow on display.

An embodiment uses forward error correction (FEC) technology to protect the specific part of video flowing so that it avoids the impact of channel error.As known in the art, produce error correction data information and be attached to the data of transmitting via communication channel such as the FEC technology of Read-Solomon and Viterbi (Viterbi).If for example, in master data (, I frame), occur wrong, FEC can be used for proofreading and correct this mistake.

FEC code increases the data rate of transmission, and therefore ideally, it only uses in the time needing most.If data are just sent out, and it will not cause very attractive visual artifacts, can goodly not carry out protected data with FEC code.For example, be next to P image block before the I image block of loss by the visual artifacts (also, the image block on screen will not be updated) that only produces 1/60 second on screen.This kind of visual artifacts almost can not be detected by human eye.Along with P image block from I image block further backward, lose P image block and all the more become more noticeable.For example, if image block circular form sample is that I image block is followed 15 P image blocks subsequently before I image block can be used again, if the P image block being next to after I image block is lost, it causes this image block to show that incorrect image lasts 15 frame times (under 60fps, this will be 250 milliseconds).Human eye is by the interruption of stream that easily detects 250 milliseconds.Therefore, P image block is apart from new I image block (also, it is more approaching that P image block is followed I image block) more backward, and false shadow is more noticeable.As discussed previously, however, but generally speaking, it is more approaching that P image block is followed I image block, less for the data of this P image block.Therefore, the P image block of following I image block not only for protection in order to avoid more crucial for losing, and its size is less.In addition, generally speaking, the data that need protection are less, protect its required FEC code less.

Therefore,, as illustrated in Figure 11 a, in one embodiment, due to the importance of I image block in video flowing, only I image block possesses FEC code.Therefore, FEC 1101 contains containing the error correcting code and the FEC 1104 that are useful on I image block 1100 error correcting code that is useful on I image block 1103.In this embodiment, do not produce FEC for P image block.

In Figure 11 b, in an illustrated embodiment, cause that for most probable in the time losing the P image block of visual artifacts also produces FEC code.In this embodiment, FEC 1105 is provided for front 3 P image blocks but is not used in the error correcting code of the P image block of following.In another embodiment, for the P image block of size of data minimum produce FEC code (its by tend to from be selected in I image block after the P image block that occurs the earliest, it is the most key for protection).

In another embodiment, not FEC code is sent together with image block, but by image block transmission twice, in different grouping, transmit at every turn.If one packet loss/delay, is used another grouping.

In Figure 11 c in a shown embodiment, produce be respectively used to and video simultaneously from the FEC code 1111 and 1113 of the audio packet 1110 and 1112 of main frame service transmission.Maintain the integrality particular importance of the audio frequency in video flowing, (for example, ticktack or fizz) will cause undesirable especially user to experience because the audio frequency of distortion.FEC code assists in ensuring that audio content reproduces at client computer 415 places undistortedly.

In another embodiment, not FEC code is sent together with voice data, but by audio data transmission twice, in different grouping, transmit at every turn.If a packet loss/delay, is used another grouping.

In addition, in Figure 11 d, in an illustrated embodiment, FEC code 1121 and 1123 is respectively used to transmit upstream to from client 415 user's input command (for example, button press) 1120 and 1122 of host services 210.This is important, may cause undesirable user to experience because miss button press or mouse movement in video-game or application program.

In another embodiment, not FEC code is sent together with user's input command data, but by user's input command transfer of data twice, in different grouping, transmit at every turn.If a packet loss/delay, is used another grouping.

In one embodiment, the quality of the communication channel of host services 210 assessments and client 415, to determine whether to use FEC, and if use what certain applications FEC of definite reply video, audio frequency and user command." quality " of assessment channel can comprise the function such as estimating packet loss, time delay etc. as above.If channel is unreliable especially, host services 210 can be to all I image blocks, P image block, audio frequency and user command application FEC.By contrast, if channel is reliable, host services 210 can be only to audio frequency and user command application FEC, or can be not to audio or video application FEC, or can not use FEC.Can use various other of the application of FEC to arrange, still observe described general principle simultaneously.In one embodiment, host services 210 constantly monitoring channel situation and correspondingly change FEC strategy.

In another embodiment, referring to Fig. 4 a and Fig. 4 b, when packet loss/delay, thereby while causing the loss of image block data, if or may make FEC can not proofread and correct the image block data of loss due to the packet loss of grain especially, client 415 be evaluated at by receive remain before new I image block how many frames and by its with from client 415 the round trip time delay to host services 210 compare.If round trip time delay is less than the number of the frame of new I image block before should arriving, client 415 sends message to host services 210, the I image block that please look for novelty.This message is routed to video compressor 404, and the P image block of its image block that not generation has been lost for data, but produce I image block.Suppose that system shown in Fig. 4 a and Fig. 4 b is through designing so that the round trip time delay that is conventionally less than 80 milliseconds to be provided, this causes image block to be corrected in 80 milliseconds (under 60fps, frame has the duration of 16.67 milliseconds, therefore in the full frame time, 80 millisecond time-delay are by the calibrated image block causing in 83.33 milliseconds, 83.33 milliseconds is 5 frame times, it is attractive interruption, but for example, far away from () noticeable for 250 milliseconds of interruptions of 15 frames).While producing this kind of I image block when compressor reducer 404 its common cyclic orders of disengaging, if the bandwidth that causes described frame is exceeded available bandwidth by I image block, compressor reducer 404 will postpone the circulation of other image blocks, to make other image blocks receive P image block (even described image duration an image block conventionally will should be I image block) during described frame time, and following common circulation will start to continue from next frame, and conventionally the image block that receives the I image block in previous frame will be received to I image block.Although the stage of this action temporary delay R frame circulation, it will be visually not noticeable conventionally.

Video and audio compressor/decompressor is implemented

Figure 12 illustrates a specific embodiment, wherein compresses concurrently 8 image blocks with multinuclear and/or multiprocessor 1200.In one embodiment, use dual core processor, four core Xeon (to strong) CPU computer system in 2.66GHz or more relative superiority or inferiority execution, each core implements to increase income x264H.264 compressor reducer as self-contained process.But, can use various other hardware/software configuration, still observe described general principle simultaneously.For example, each in core cpu can be replaced by the H.264 compressor reducer of implementing with FPGA.In Figure 12, in shown example, core 1201-1208 is for process I image block and P image block as eight independent lines threads simultaneously.As well-known in technique, current multinuclear and multiprocessor computer system and when processing operating system and integrate such as the multi-thread thread of Microsoft Windows XP professional version (64 than the special edition of a book or 32 than the special edition of a book) and Linux, it can carry out multi-thread thread processing inherently.

In embodiment illustrated in fig. 12, because each in these 8 cores is only responsible for an image block, so it is independent of to a great extent other cores and operates, each carries out the independent instantiation of x264.Use taking PCI Express x1 as basic DVI capured keystroke (such as, from the Sendero video imaging IP development board of the Microtronix of Oosterhout of Netherlands) catch the video of the uncompressed under 640 × 480,800 × 600 or 1280 × 720 resolution, and FPGA on card uses direct memory access (DMA) (DMA) that caught video is sent in system RAM via DVI bus.Described image block is configured to 4 × 2 configuration 1205 (although it is illustrated as square image block, in this embodiment, they have 160 × 240 resolution).Each instantiation of x264 is configured to compress the one in these 8 160 × 240 image blocks, and it enters a circulation through synchronization with each core after making to compress at initial I image block, each frame and not homophase of another frame, with compress that an I image block continues with seven P image blocks, as illustrated in Figure 12.

At each frame time, use previous described technology that the compressed image block of gained is combined into grouping stream, and then compressed image block is transferred to destination client 415.

Although undeclared in Figure 12, if exceeding, the data rate of 8 image blocks of combination specifies peak data rate 941, all 8 x264 processes reach supspend to last at necessary frame time, until transmitted the data of 8 image blocks for combining.

In one embodiment, client 415 is embodied as to the software on the PC of 8 instantiations that carries out FFmpeg.Receiving course receives 8 image blocks, and each image block is routed to FFmpeg instantiation, and FFmpeg instantiation decompressing image piece is also reproduced to the suitable image block position on display device 422.

Client 415 receives from the keyboard of the input equipment driver of PC, mouse or game console input and is transferred to server 402.The server 402 input equipment data that then application receives are also applied to game or the application program on server 402, carried out, and server 402 is for using Intel 2.16GHz double-core CPU to carry out the PC of Windows.Server 402 then produces new frame and via its DVI output, new frame is exported certainly taking motherboard as basic graphics system or via the DVI output of NVIDIA8800GTX pci card.

Simultaneously, server 402 via its digital audio output (for example, S/PDIF) output is by the audio frequency of playing or application program produces, and this digital audio output is coupled to the digital audio input on basic PC taking two four core Xeon of implementing video compression.Vorbis increases income audio compressor for whatsoever core and side by side compressed audio of video with can be used for processing line thread.In one embodiment, first the core that completes its image block of compression carries out audio compression.Then compressed audio frequency is transmitted together with compressed video, and in client 415 with the Vorbis audio decompressor compressed audio frequency that decompresses.

Host services server-centric distributes

Light via glass (such as, optical fiber) is advanced with a certain part of light speed in a vacuum, and therefore can determine the definite propagation velocity of light in optical fiber.But, in practice, considering the time for routing delay, transmission inefficiency and other consumptions, we observe optimum delay reflection on the Internet and approach 50% transmission speed of the light velocity.Therefore, best 1000 miles of round trip time delays are approximately 22 milliseconds, and best 3000 miles of round trip time delays are approximately 64 milliseconds.Therefore, the individual server on U.S.'s seashore consequently can not be with the client (it may reach 3000 miles) on servo another seashore of desired time delay by hypertelorism.But, as illustrated in Figure 13 a, for example, if the center that host services 210 server-centrics 1300 are positioned the U.S. (, the Kansas State, the Nebraska State etc.), so that to the distance of any point in ConUS be approximately 1500 miles or below 1500 miles, the time delay of round trip the Internet can be low to moderate 32 milliseconds.Referring to Fig. 4 b, note: although the time delay of situation worst that user ISP 453 allows is 25 milliseconds, conventionally, the in the situation that of DSL and cable modem system, we observe the time delay that approaches 10-15 millisecond.And from customer site 211, the ultimate range to Co-location center 210 is 1000 miles in Fig. 4 b supposition.Therefore, in the case of the used user ISP round trip time delay of typical 15 milliseconds and the maximum the Internet distance of 1500 miles for the round trip time delay of 32 milliseconds, the moment that activates input equipment 421 from user is to seeing that on display device 422 total round trip time delay of response is 1+1+15+32+1+16+6+8=80 millisecond.Therefore, conventionally can in the Internet distance of 1500 miles, reach 80 milliseconds of response times.This is accessed in any customer site enough in permission ConUS with short user ISP time delay 453 at the individual server center of centralized positioning.

In Figure 13 b in illustrated another embodiment, in host services 210 server-centric HS1-HS6 strategies, be positioned the U.S. (or other geographic areas) around, specificly approach high population center and locate (for example, HS2 and HS5) compared with large host services server-centric.In one embodiment, server-centric HS1-HS6 is via network 1301 exchange messages, and network 1301 can be the Internet or private network or both combinations.The in the situation that of multiple server-centric, can provide service to the user with high user ISP time delay 453 compared with low delay.

Although the distance on the Internet is to the influential factor of round trip time delay via the Internet really, sometimes also work with irrelevant other factors of time delay to a great extent.Sometimes via the Internet, stream of packets routed to the position of distance and again returns, thereby causing from macrocyclic time delay.Sometimes on path, there is the routing device of inappropriate operation, thereby cause the delay of transmission.Sometimes have the communication that makes path overload, it introduces delay.In addition sometimes, be to exist to prevent that user's ISP from routing to the fault of given destination, at all.Therefore, although common interconnection network provides the connection from any to another point with quite reliable and best route and time delay conventionally, the route that this is quite reliable and best and time delay are to determine (especially in the case of causing the outside long distance join of the local zone that routes to user) by distance to a great extent, but this reliability and time delay can not get any guarantee and usually can not be from user's place to the given destination on common interconnection network and reach.

In one embodiment, in the time that subscription client 415 is connected at first host services 210 with playing video game or uses application program, client in the time starting with each communicate by letter (for example, using technology as described above) in available host services server-centric HS1-HS6.If time delay is enough low for specific connection, use this connection.In one embodiment, client is communicated by letter with a subset of All hosts service server center or host services server-centric, selects to have the host services server-centric that minimum time delay connects.Client can select to have the service centre that minimum time delay connects, or server-centric can be identified and has server-centric that minimum time delay connects and this information form of internet address (for example, with) is offered to client.

If particular host service server center overload and/or user's game or application program tolerable to another, the time delay of the less host services server-centric being written into, client 415 can be redirected to another host services server-centric.Under this kind of situation, the game that user is just being carried out or application program are suspended on the server 402 at user's overload server-centric place, and game or application state data are sent to the server 402 at another host services server-centric place.Then will restart this game or application program.In one embodiment, host services 210 by wait until game or application program reach nature breakpoint (for example, between the rank in game, or user's initial in application program " preservations " operate after) just transmit.In another embodiment, host services 210 by wait for example, until till User Activity stops lasting cycle fixed time (, 1 minute) and then by this moment initial transmission.

As mentioned above, in one embodiment, host services 210 is subscribed to the internet bypass service 440 of Figure 14 to attempt that guaranteed time delay is offered to its client.Internet bypass service is to provide any for example, service to the private network route of the guaranteed characteristic of having of another point (, time delay, data rate etc.) on the Internet as used herein.For example, if host services 210 is just being used the DSL service reception of the AT & T providing in San Francisco from user's mass communication (but not routing to the central office as base taking the San Francisco of AT & T), host services 210 will be rented from ISP's's (may be AT & T itself or another supplier) high power capacity private data and be connected between one or many person of the central office taking San Francisco as base and the server-centric for host services 210.Then,, if use the user's of AT & T DSL route to cause too high time delay via common interconnection network from All hosts service server center HS1-HS6 to San Francisco, can change into and use private data to connect.Although private data connects conventionally more expensive than the route via common interconnection network, as long as it keeps a little percentage of host services 210 to be connected to user, total cost impact is just low, and user will experience more consistent service experience.

The in the situation that of power failure, server-centric usually has two non-firm power layers.Ground floor is generally the non-firm power from battery (or from the energy source that can use immediately substituting, such as keeping running and being attached to the flywheel of generator), and it provides immediately electric power and keeps server-centric running in the time that feeder cable is out of order.If power failure is temporary transient, and feeder cable returns to rapidly (for example, in one minute), and what battery was required is to keep server-centric running.If but power failure is lasted the longer time cycle, conventionally starter-generator (for example, diesel engine power supply) as long as replace battery and generator has fuel and can turn round.Described power generator electrode costliness, because it must produce the electric power that nearly server-centric obtains from feeder cable conventionally.

In one embodiment, each in host services HS1-HS5 shares user data each other, so that in the time that a server-centric has power failure, it can suspend afoot game and application program, and then game or application state data are sent to the server 402 at other server-centric places from each server 402, and then will notify each user's client 415 to instruct it to be conveyed to new server 402.Suppose that described situation occurs once in a while, user is transferred to and can not provides the host services of optimum delay server-centric (to be also, user will only must tolerate that higher time delay lasts the duration of power failure) can be acceptablely, it will be allowed for the option of much wide scope of transferring user.For example, the time zone of the given leap U.S. is poor, and the user on east coast may will sleep at 11:30PM, and user on West Coast is just starting to reach peak value in video-game use at 8:30PM.If there is power failure in the host services server-centric at that time on West Coast, may not there is not the enough West Coast servers 402 for the treatment of all users in other host services server-centric places.Under this kind of situation, some users can be transferred to the host services server-centric on east coast with available server 402, and unique consequence for user will be higher time delay.Once user is shifted from the server-centric that loses electric power, and server-centric can start the orderly cut-out of its server and equipment with that, to cut off all devices before battery (or other immediately CAPACITY IN ELECTRICITY) exhausts.In this way, can avoid the cost for the generator of server-centric.

In one embodiment, at the time durations being seriously written into of host services 210 (or because peak user is written into, or because one or more server-centrics are out of order), user is transferred to other server-centrics by the game just using based on user or the delay requirement of application program.Therefore, by for use need that the game of low delay or the user of application program provide that available low delay server to there is limited supply connects preferably.

Host services feature

The embodiment of the assembly of the server-centric for host services 210 that Figure 15 explanation utilizes in following characteristics is described.As host services illustrated in Fig. 2 a 210, unless have ready conditions in addition, the assembly of this server-centric is controlled and is coordinated by host services 210 control system 401.

The port internet communication 1501 that enters from subscription client 415 is directed to into port route 1502.Conventionally, enter port internet communication 1501 high speed fibre via to the Internet is connected and enters server-centric, but any network connection device with enough bandwidth, reliability and low delay will be enough.Entering port route 1502 is that (this network can be embodied as Ethernet, fibre channel networks to network, or via any other conveying device) interchanger and support the system of the routing server of described interchanger, its obtain the grouping of arrival and by each Packet routing to appropriate application/game server 1521-1525.In one embodiment, the grouping that is sent to application-specific/game server represent the data that receive from client a subset and/or can be by other assemblies in data center (for example, network coupling assembling, such as gateway and router) translate/change.Under some situations, (for example) if game or application program on multiple servers, carry out concurrently simultaneously, at every turn by more than Packet routing to server 1521-1525.RAID array 1511-1512 is connected to into port route network 1502, to make application program/game server 1521-1525 can read RAID array 1511-1512 and to write RAID array 1511-1512.In addition, RAID array 1515 (it can be embodied as multiple RAID arrays) is also connected to into port route 1502, and can read by self-application program/game server 1521-1525 from the data of RAID array 1515.Entering port route 1502 can implement in the multiple prior art network architecture (comprise the interchanger of tree structure, enter port internet communication 1501 at its root); In the network structure of all various devices of interconnection, implement; Or implement as the sub-network sequence (the concentrated communication isolating in the middle of concentrated communication and other equipment in the middle of interwork unit) of interconnection.The network configuration of one type is SAN (storage area network), although it is generally used for storage facilities, it also can be used for the universal high speed data transmission between equipment.Again, application program/game server 1521-1525 can have separately to the multiple networks connections that enter port route 1502.For example, server 1521-1525 can have to the network of sub-network that is attached to RAID array 1511-1512 and connects and connect to another network of the sub-network that is attached to other equipment.

Application program/game server 1521-1525 can through in the same manner, some differently or all differently configures, as previously described about the server 402 in embodiment illustrated in Fig. 4 a.In one embodiment, each user is generally at least one application program/game server 1521-1525 in the time using host services.For the purpose of explanation simple, will suppose that given user is just using application program/game server 1521, but multiple server can use by a user, and multiple user can share single application program/game server 1521-1525.The user's who sends from client 415 (as described previously) control inputs is through being received as into port internet communication 1501, and routes to application program/game server 1521 via entering port route 1502.Application program/game server 1521 users' control inputs is as to the game carried out on server or the control inputs of application program, and calculates the next frame of video and audio frequency associated with it.Application program/game server 1521 then exports the video/audio of uncompressed 1529 to and shares video compression 1530.Application program/game server can be exported via any device (comprising that one or more ultrahigh speed Ethernet connects) video of uncompressed, but in one embodiment, video is connect and export via DVI (Digital Video Interactive), and audio frequency and other compressions and communication channel state information system connect and export via universal serial convergence bus (USB).

Share video compression 1530 and compress video and the audio frequency from the uncompressed of application program/game server 1521-1525.This compression can be implemented with hardware or with the hardware of executive software completely.Can there is the special compressor reducer for each application program/game server 1521-1525, if or compressor reducer enough fast, can compress the video/audio from more than one application program/game server 1521-1525 with given compressor reducer.For example, under 60fps, video frame time is 16.67 milliseconds.If compressor reducer can compress 1 frame in 1 millisecond, this compressor reducer can be used for compressing from the video/audio that reaches 16 application program/game server 1521-1525 by obtaining from the input of server one by one, and this compressor reducer is preserved the state of each video/audio compression process and worked as its circulation time in the middle of the video/audio stream from server and switches background.This causes the essence cost savings of compression hardware.Because different server will complete frame at different time, so in one embodiment, compressor reducer resource be shared storage device in thering is the state for storing each compression process (for example, RAM, flash memory) shared collection district 1530 in, and when server 1521-1525 frame is complete and prepare when compressed, control device determines which compressed resource can be used at that time, for this compressed resource provides the state of compression process of server and the frame of the video/audio of uncompressed to be compressed.

Note, comprise the information about compression itself for a part for the state of the compression process of each server, such as the frame buffered data (it can be used as the reference for P image block) through decompressing of previous frame, the resolution that video is exported; The quality of compression; Image block structure; The distribution of the bit of every image block; Compression quality, audio format (for example, stereo, around audio, aC-3 (Doby ) AC-3).But compression process state also comprises about following communication channel state information: peak data rate 941, and current whether being just output of previous frame (as illustrated in Fig. 9 b) (and therefore should ignore present frame), and whether exist potentially should in compression, consider (such as, too much packet loss) impact compression decision-making is (for example, aspect the frequency of I image block, etc.) the characteristic of channel.Because peak data rate 941 or other characteristics of channel changed along with the time, as by supporting that each user monitors that application program/game server 1521-1525 of the data that send from client 415 is determined, so being sent to common hardware by application program/game server 1521-1525, relevant information compresses 1530.

Common hardware compression 1530 is also used the device of all described devices as described earlier by compressed video/audio packetizing, and in due course, application FEC code, copy particular data, or take other steps, to guarantee fully the ability that video/audio data stream is received by client 415 and decompress with feasible high-quality and reliability.

Some application programs (such as, below described application program) need the video/audio output while under multiple resolution (or with other multiple forms) of given application program/game server 1521-1525 available.If application program/game server 1521-1525 so notifies common hardware to compress 1530 resources, the video/audio 1529 of the uncompressed of this application program/game server 1521-1525 will be by with different-format, different resolution and/or compression simultaneously in different grouping/error correction structure.Under some situations, some compressed resources can share and (for example,, in many compression algorithms, exist so as to the front step that image is adjusted in proportion to multiple sizes in applied compression in the middle of multiple compression processes of the same video/audio of compression.If desired the image of the different sizes of output, this step can be used for serving several compression processes simultaneously).Under other situations, for each form by compressed resource independent needs.Under any situation, the compressed video/audio 1539 that is used for the required all various resolution of given application program/game server 1521-1525 (one or more) and form is outputed to the route 1540 that goes out of the port simultaneously.In one embodiment, the output of compressed video/audio 1539 is in UDP form, and therefore it is unidirectional stream of packets.

The route network 1540 that goes out of the port comprises a series of routing servers and interchanger, this series routing server and interchanger are directed to each compressed video/audio stream user or other destinations of having a mind to and/or are back to delay buffer 1515 via internet communication 1599 interfaces that go out of the port (it is conventionally by the optical fiber interface being connected to the Internet), and/or be back to into port route 1502, and/or export for carrying out video distribution via private network (not shown).Note (as described below): the route that goes out of the port 1540 can export given video/audio stream to multiple destinations simultaneously.In one embodiment, this is to use Internet protocol (IP) multicast to implement, and wherein broadcast is intended to flow to the given UDP stream of multiple destinations simultaneously, and this broadcast carrys out repetition by the routing server and the interchanger that go out of the port in route 1540.The plurality of destination of broadcast can arrive via the Internet multiple users' client 415, via entering port route 1502 to multiple application program/game server 1521-1525, and/or arrive one or more delay buffers 1515.Therefore, the output squeezing of given server 1521-1522 is become to one or more forms, and each compressed stream is directed to one or more destinations.

In addition, in another embodiment, for example, if multiple application program/game server 1521-1525 (are used by a user simultaneously, for generation of having in the parallel processing configuration of 3D output of complex scene) and each server produce the part of gained image, can compress 1530 outputs of the video by multiple server 1521-1525 by common hardware and be combined into a combined frames, and this point is processed forward this combined frames as mentioned above certainly, seem that it is from single application program/game server 1521-1525.

Note, in one embodiment, the duplicate of all videos that produced by application program/game server 1521-1525 (at least taking the resolution of the video watched by user or high-resolution more) is recorded in to minute (in one embodiment as 15 minutes) of in delay buffer 1515, lasting at least a certain number.This allows the video of each user " playback " from each session, to verify previously work or achievement (under the situation of game).Therefore, in one embodiment, each compressed video/audio output 1539 streams that route to subscription client 415 are also multicasted to delay buffer 1515.In the time video/audio being stored on delay buffer 1515, on the network address that the catalogue on delay buffer 1515 provides application program/game server 1521-1525 (its source of video/audio for postponing) and delay buffer 1515, can find the cross reference between the position of video/audio of delay.

Game live, that can at once watch, that can at once play

Application program/game server 1521-1525 not only can be used for carrying out user's given application program or video-game, and it can be used for setting up for supporting the user interface application via the navigation of host services 210 and the host services of other features 210.A kind of screen shot of this user interface application is presented at (" game view finder (Game Finder) " screen) in Figure 16.This particular user interface screen allows user to watch on-the-spot (or the delay) of playing 15 game by other users.In " thumbnail " video window each (such as, 1600) be live video window at the volley, it shows a video from a user's game.In thumbnail, shown view can be the same view that user is just seeing, or its view that can be delay (for example, if user is just playing fight game, user may not wish that other users see where it be hidden in and it can select any view of its game play to be postponed to a time cycle (for example, 10 minutes)).View also can be the camera view of the game of the view that is different from any user.By menu setecting (not shown in this explanation), user can select based on multiple standards the selection of the game that will simultaneously watch.As the little sampling of exemplary selection, user can select the player of the given rank in top player, the game of the game (playing by different players) of random selection (all game as shown in Figure 16), all classifications of game, only game, or more rudimentary player (for example,, if the positive learning foundation of player), be " partner " (or for competitor) player, there is maximum number beholders' game etc.

Note, conventionally, whether each user can be watched by other people from the video of its game or application program determining, and if so, determine that this video can other people be watched and when can be watched by other people by which, determine whether this video is watched only in the situation that having delay.

Application program/game server the 1521-1525 that produces user interface screen shown in Figure 16 obtains these 15 video/audios by the application program/game server 1521-1525 transmission message to each user (this application program/game server 1521-1525 is just asking the game from this user) and is fed to.This message is via entering port route 1502 or another network sends.This message will comprise size and the form of requested video/audio, and identification is watched to the user of user interface screen.Given user can select " piracy " pattern and disapprove the video/audio (watch a little or from another and watching a little from it) that any other user watches its game, or as previous section described in, user can select to allow to watch the video/audio from its game, but postpones the video/audio of watching.Therefore user application/game server 1521-1525 (its reception also accepts to allow the viewed request of its video/audio) will confirm to request server, and it also will notify common hardware compression 1530 need to produce the extra compressed video flowing of requested form or screen size (supposition form and screen size are different from the form and the screen size that have produced), and it is also by the destination (also, request server) of the compressed video of instruction.If requested video/audio is only delayed, request applications/game server 1521-1525 will be notified like this, and it obtains the video/audio of delay by the position by searching the video/audio in the catalogue on delay buffer 1515 and for the network address self-dalay buffer 1515 of application program/game server 1521-1525 in the source of the video/audio that postpones.Once all these requests are produced and are processed, the video flowing up to 15 live thumbnail size is routed to into port route 1502, to the application program/game server 1521-1525 that produces user interface screen from the route 1540 that goes out of the port, and will will decompress and show by this server.The video/audio stream postponing may be in excessive screen size, if like this, application program/game server 1521-1525 is by described decompression stream and video flowing is reduced to thumbnail size in proportion.In one embodiment, serve (in Figure 15, not showing) (and being managed by central authorities' " management " service) by the request of audio/video being sent to the host services control system similarly central " management " of Fig. 4 a, central authorities' " management " serve is then redirected to appropriate application/game server 1521-1525 by described request.In addition, in one embodiment, may not need request, because thumbnail is " pushed " to those users' the client that allows it.

All audio frequency that simultaneously mix from 15 game may produce ear-piercing sound.User can select in this way all sound mix (may be just the sensation that obtains " clamour " that produced by viewed everything) together, or user can select only to listen to the audio frequency from a game at every turn.The selection of single game be by yellow is selected frame 1601 move to given game complete (yellow frame move can by the arrow key with on keyboard, by rolling mouse, by mobile operating bar or by promote complete such as the arrow button on another equipment of mobile phone).Once select single game, only play from the audio frequency of this game.And, show game information 1602.Under the situation of this game, for example, publisher's mark (" EA ") and game mark " this valley of Need For Speed card " and orange horizontal stripe are indicated in this particular moment and are played games or watch the people's of game number (under this situation under relative conditon, many, therefore game is " hot topic ")." state " is provided in addition, 80 different instantiations that instruction exists 145 players to play just energetically Need For Speed game (are also, this game can be played by indivedual player or multiplayer), and there are 680 beholders (this user is one of them).Note, it is upper that RAID array 1511-1512 is collected and be stored in to this statistics (and other statisticss) by host services control system 401, for keeping daily record that host services 210 operates and for suitably to user's charging and to publisher's defrayment that content is provided.Some statisticss are records due to the action of being undertaken by operation control system 401, and some statisticss are to report to operation control system 401 by indivedual application program/game server 1521-1525.For example, in the time that game is just viewed (and being stopped while watching when playing), application program/game server the 1521-1525 that carries out this game view finder application program sends message to host services control system 401, so that obtain the statisticss of renewable how many game of main frame operation control system 401 in watching.It is used that some statisticss can be user interface application (such as, this game view finder application program).

If the start button on its input equipment of user click, it will see that the thumbnail video amplifier thumbnail video maintenance simultaneously on-the-spot broadcasting in yellow frame is full screen size.This effect is presented in the process in Figure 17.Note, the size of video window 1700 increases.In order to implement this effect, application program/game server 1521-1525 has the duplicate of the video flowing of the full screen size (with the resolution of user's display device 422) of the game that is routed to it from moving application program/game server 1521-1525 request of selected game.Application program/game server 1521-1525 notice common hardware compressor reducer 1530 of carrying out game no longer needs the duplicate (unless this thumbnail of Another Application program/game server 1521-1525 needs) of the thumbnail size of game, and then it guides common hardware compressor reducer 1530 duplicate of the full screen size of video to be sent to application program/game server 1521-1525 of amplified video.The user who plays this game can or can not have the display device 422 that resolution is identical with the resolution of the user's that game is amplified described display device.In addition, other beholders of game can or can not have resolution and the identical display device 422 of user that game is amplified (and can there is different audio playbacks, for example, stereo or around audio).Therefore, common hardware compressor reducer 1530 determines whether to produce the suitable compressed video/audio stream of the user's who meets request video/audio stream requirement, if and suitable compressed video/audio stream exists really, common hardware compressor reducer 1530 notifies the route 1540 that goes out of the port the duplicate of this stream to be routed to the application program/game server 1521-1525 that amplifies this video, if and suitable compressed video/audio stream does not exist, another duplicate that is suitable for described user of compressed video and guides the route that goes out of the port that this stream is sent back to into port route 1502 and amplifies application program/game server 1521-1525 of this video.This server of the present full screen version that receives selected video is by this full screen version of decompression and it is little by little amplified to full size in proportion.

How Figure 18 explanation screen after game being amplified to completely to full screen and showing game with the full resolution of user's display device 422 seems (as indicated in the image pointing to by arrow 1800).Application program/game server the 1521-1525 that carries out game view finder application program sends message to other application program/game servers 1521-1525 that thumbnail is provided and no longer watches other game to indicate described thumbnail no longer to need and to send message to host services Control Server 401 with instruction.Now, unique superimposed layer (overlay) 1801 that is shown as the top of screen of generation, information and menu control are offered user by it.Note, along with this game progress, spectators rise to 2,503 beholders.In the case of so many beholder, certainly exist many beholders with display device 422, described display device 422 has identical or approaching resolution (each application program/game server 1521-1525 have adjust in proportion video for adjusting the ability of fitness).

Because shown game is multiplayer, so user can determine to carve at a time and add this game.Due to many reasons, host services 210 can or can not allow user to add this game.For example, user may must pay the expense of playing games and select not pay, user may not have (is for example enough to add enough grades of described particular game, for other players, it will be not competitive), or user's the Internet connects not have is enough to allow that user plays that enough low time delay is (for example, there is not the time delay constraint for watching game, therefore can in the situation that paying close attention to without time delay, watch by a distance (in fact, land greatly at another) play game, but for game to be played, time delay must be enough low so that user (a) enjoys this game, and (b) in the status equal with thering are other players of being connected compared with low delay).If permitting user plays, for user provides application program/game server 1521-1525 of game view finder user interface by 401 initialization of requesting host service control server (be also, locate and start) thus the application program/game server 1521-1525 being configured to suitably for playing particular game is written into this game from RAID array 1511-1512, and then host services Control Server 401 switches to compression video/audio from the application program/game server that to play carry out Co-location now by be sent to now application program/game server that game is carried out the application program/game server of Co-location and now game carried out to Co-location from user's control signal by the video/audio that instructs common hardware compression 1530 application program/game server from Co-location game view finder application program from compression by instructing into port route 1502.Game view finder application program/game services with game is carried out to the vertical synchronization of new application program/game server of Co-location asynchronous, and therefore may life period poor between synchronous at these two.To after completing frame of video, application program/game server 1521-1525 start compressed video because share video compression hardware 1530, so complete and to obtain early from the full frame time of the comparable legacy server of the first frame of new server, the first frame from new server may be before previous compressed frame completes its transmission (for example, consider the transmission time 992 of Fig. 9 b: if the frame of uncompressed 3963 early a frame time half complete, it will affect (impinge) transmission time 992).In this case, sharing video compression hardware 1530 will from the first frame of new server (for example ignore, as ignore (974) frame 4964), and the most end frame from legacy server is kept an extra frame time by client 415, and shared video compression hardware 1530 will start compression from the next frame time video that game is carried out to the new application program/game server of Co-location.For user, visually, the transformation from an application program/game server to Another Application program/game server will be seamless.Application program/game server 1521-1525 that host services Control Server 401 then carries out Co-location by notice to game view finder switches to idle state, until again need it.

User then can play this game.In addition, what make an exception is that game will be play immediately (because game is loaded into application program/game server 1521-1525 from Raid array 1511-1512 with 1,000,000,000 bps of speed) in perception, and the driver by desirable, buffer configuration (under the situation of Windows) will be played together with being correctly configured to be loaded into for the operating system of this game the server that is accurately suitable for this game, and do not have other application programs of the operation competition of possibility and this game to carry out on this server.

And, along with user makes progress in game, each in the fragment of game will with the speed of 1,000,000,000 bps (be also, in 8 seconds, be written into gigabyte) be written into server from RAID array 1511-512, and due to huge storage volume (because its common source that is many users of RAID array 1511-1512, so it may be very large, but still cost-effective), make computational geometry shape setting in advance or other game fragment settings and be stored in RAID array 1511-1512 above and be extremely fast written into.In addition, because the hardware configuration of each application program/game server 1521-1525 and computing capability are known, so calculating pixel and vertex coloring in advance.

Therefore, game can almost start at once, and it will carry out in ecotopia, and fragment subsequently will almost be written at once.

But except these advantages, user can watch other people to play games by (via previous described game view finder, and other device), and both all determine whether game is interesting, if like this, from watching other people to carry out learning skill.In addition, user can demonstrate this game immediately, and needn't wait for large download and/or installation, and user can play this game (may on the basis on probation of using compared with tip, or on longer basis) at once.In addition, user can be by the wireless connections of enough low delays on Windows PC, Macintosh, on television set, at home, while advancing and even play this game on mobile phone.In addition, this all can complete in the situation that not once entity has game duplicate.

Narrate as previous, user can determine not allow its game play to be watched by other people, allows its game after postponing, to watch, and allows its game can be selected user and watches, or allow its game to be watched by all users.In any case, in one embodiment, video/audio is stored in delay buffer 1515 and lasts 15 minutes, and user can " return " and watch its previous game play, and by suspension of game play, the playback lentamente of playing, the F.F. etc. of playing, can carry out in the time watching the have digital video recorder TV of (DVR) as it.Although in this example, user is playing games, if user is just using application program, identical " DVR " ability is available.This is verifying in previously work and useful in other application as detailed below.In addition, if game is returned the ability of falling through being designed to have based on utilizing game state information, to can change camera view etc., also will support this " 3D DVR " ability, but it need to be by game design for supporting " 3D DVR " ability." DVR " ability of use delay buffer 1515 will be (certain together with any game or application program, be limited to the video producing in the time using game or application program) work together, but under the situation of game with 3D DVR ability, the 3D that user can control the fragment of previously playing " leaps (fly-through) ", and makes delay buffer 1515 record the game state of gained video recording game fragment.Therefore, specific " leaping " is recorded as to compressed video, but because also by recording game state, so different leaping may be on the date after a while of the same fragment of game.

As described below, the user in host services 210 will have a User Page separately, and in this User Page, user can announce about the information of itself and other data.The video segment that one of thing that user can announce is the game play of having preserved from it.For example, if user has overcome the challenge of difficulty especially in game, before user can just " return " and obtain the place of its large achievement to it in game, and then instructing host services 210 video segment of a certain duration (for example, 30 seconds) to be kept on user's User Page watches for other users.In order to implement these, the thing that application program/game server 1521-1525 that user is just using only will do is to be stored in video playback in delay buffer 1515 to RAID array 1511-1512 and then on user's User Page, to index to described video segment.

If game has the ability of 3D DVR, as mentioned above, the User Page that also can be recorded the required game state information of 3D DVR and be user by user can be used.

In game through being designed to also there is " onlooker " except thering is the player of enlivening (be also, can in the situation that not participating in, advance and observe the user of action in the 3D world), the view finder application program of playing will make user add game as onlooker and player.From the enforcement point of watching, for host computer system 210, user is onlooker instead of enlivens player and not there are differences.Game is written into the upper and user of application program/game server 1521-1525 and will controls this game (for example, the virtual camera in the world is watched in control).Unique difference is user's game experiencing.

Multiple users cooperate

Another feature of host services 210 is the ability (even if watching also like this with far different equipment) that multiple users cooperate in watching live video.In the time playing games and in the time using application program, this is all useful.

Many PC and mobile phone are equipped with video camera and have the ability (especially when image hour) of carrying out Real-time Video Compression.And little camera is available, can be attached to TV, and with software or not difficult with implementing Real Time Compression for the one of many hardware-compressed equipment of compressed video.And many PC and all mobile phones have microphone, and earphone have in microphone situation available.

Combination has described camera and/or the microphone of local video/audio compression ability (specifically, using low delay video compression technology described herein) will make user video and/or audio can be transferred to host services 210 together with input device controls data from customer site 211.When using when described technology, can realize ability illustrated in fig. 19: user can make its video and audio frequency 1900 appear on the screen in another user's game or application program.This example is multiplayer, and wherein teammate cooperates in racing car.User's video/audio only can optionally be seen and see/hear by its teammate.In addition, because use technology as described above effectively not have time delay, so player can talk each other in real time or move and there is no appreciable delay.

It is to have arrived as entering port internet communication 1501 from the compressed video and/or audio of camera/microphone of user by making that this video/audio is integrated.Then, entering port route 1502 is routed to this video and/or audio to be allowed to see the application program/game server 1521-1525 that see/hear video and/or audio.Then, user's decompressed video and/or the audio frequency of application program/game server 1521-1525 out of the ordinary of choice for use video and/or audio and optionally integrated to come across in game or application program, such as by 1900 illustrated.

The example of Figure 19 shows how in game, to use this cooperation, but this cooperation can be the extremely powerful instrument for application program.Consider each situation: wherein an edifice just by the architect in Chicago for taking New York as the real estate developer in base is for designing New York, but this decision-making relates in tourism and happens to be in the financial investment person on Miami airport, and need to carry out the decision-making of the particular design key element (aspect near the building how it arranges in pairs or groups it) about building, to meet investor and real estate developer.Suppose that building constructor has the high resolution monitor of the camera that is attached to PC in Chicago, real estate developer has the notebook with camera in New York, and investor has the mobile phone with camera in Miami.Building constructor can carry out Co-location to the powerful architectural design application program that can carry out high fidelity 3D reproduction by host services 210, and it can utilize the large database concept of the building of New York, and the database of the building just designing.Architectural design application program by the one in application program/game server 1521-1525 (if or its need a large amount of computing capabilitys, on some persons) carry out.Each in 3 users in disparate (disparate) position will be connected to host services 210, and each time is watched the video output having architectural design application program, but it is suitably set size (for example to locking equipment and network connection performance by common hardware compression 1530 by what have for each user, building constructor can connect and see that 2560 × 1440 60fps show via the commercial the Internet of 20Mbps, the real estate developer in New York can connect and see 1280 × 720 60fps images via the 6Mbps DSL on its notebook, and investor can connect and see 320 × 180 60fps images via the 250Kbps honeycomb fashion data on its mobile phone).Each party will hear the voice (any one in the many widely available Conference Calling package software by application program/game server 1521-1525 processed to Conference Calling) of its other party, and via the actuating of the button on user input device, user can use its this ground camera that video is occurred.Along with meeting is carried out, architect can reproduce to show what building looks like in the time that it makes building rotation and makes it leap (fly-by) in abutting connection with another building in this region by having the 3D of photo realistic sense, and all Fang Jun will can see same video under the resolution of the display device in each side.Any one in the local device being used by any side can not be all problem with this realistic processing 3D animation, says nothing of and downloads or even store the required huge database of Adjacent Buildings that reproduces New York.The viewpoint of each in user is seen, although distance is far, although and be disparate local device, it will have seamless experience simply under fabulous sense of reality degree.In addition,, in the time that a side wishes that its face be it seems and preferably passes on its emotional state, it can so carry out.In addition, if real estate developer or investor wish to control building construction program and use the input equipment of himself (it is keyboard, mouse, keypad or touch screen), it can be like this, and it can respond with the time delay of unaware (supposing that its network connection does not have irrational time delay).For example, under the situation of mobile phone, if mobile phone is connected to the WiFi network on airport, it will have low-down time delay.If but its cellular data network that uses the U.S. can use now, it probably will suffer attractive hysteresis.But for most of objects of meeting (wherein investor is just watching architect to control that building leaps or just talking about video conference call), even honeycomb fashion time delay should be also acceptable.

Finally, in the time that cooperative Conference Calling finishes, real estate developer and investor will carry out its comment and break from host services, and the video that has been recorded in the meeting on delay buffer 1515 and comment, facial expression and/or the action of verifying the 3D model that is applied to building carrying out in the session can " be returned " by building constructor.If exist it to wish the specific fragment of preserving, the described fragment self-dalay buffer 1515 of video/audio can be moved to RAID array 1511-1512 and store and playback after a while for archives.

And, from cost viewpoint, if architect only needs to use the computing capability of New York and the Conference Calling that large database concept lasts 15 minutes, it only needs to pay the expense of the time that described resource used, and needn't have the workbench of high ability and needn't buy the expensive duplicate of large database concept.

The community service that video is abundant

Host services 210 makes unprecedented chance set up on the internet the abundant community service of video.Figure 20 shows the exemplary User Page for the game player in host services 210.As game view finder application program, User Page is the application program of carrying out in the one in application program/game server 1521-1525.All thumbnails on this page and video window show mobile consistently video (if fragment is short, its circulation).

Use video camera or pass through uploaded videos, user's (its user is called " KILLHAZARD ") can announce the video 2000 (other users can watch this video) of itself.This video storage is on RAID array 1511-1512.And, in the time that other users come the User Page of KILLHAZARD, if KILLHAZARD is now just using host services 210, its live video 2001 whatsoever just carrying out (supposition KILLHAZARD permits watching the user of its User Page to watch this video) will be shown.Whether this asks KILLHAZARD by the application program/game server 1521-1525 by User Page application program being carried out to Co-location from operation control system 401 is that (and if so, the application program/game server 1521-1525 that asks it just using) enlivened completes.Then, use the same procedure being used by game view finder application program, the compressed video flowing of adequate resolution and form is sent to and carries out application program/game server 1521-1525 of User Page application program and by its demonstration.If user selects the window of the live game play with KILLHAZARD and then suitably clicks its input equipment, this window will amplify (reusing and the identical method of view finder application program of playing), and the resolution of the display device to watch user 422 (being suitable for the characteristic of the Internet connection of watching user) is filled screen by live video.

This key advantage that is better than prior art approach is: the game of watching on-the-spot broadcasting that the user of User Page can see that user does not have and playing, and can not there is local computer or the game console that can play this game.It sees in User Page and is shown as the fabulous chance that the user of " in activity " plays games for user provides, and this is the chance of understanding the game of watching user may wish to attempt or to be good at.

Video clipping cameras record or that upload from the partner 2002 of KILLHAZARD is also shown on User Page, and the below of each video clipping is this partner of instruction word of whether playing games online (for example, six_shot just play games " Dragon Knight (Eragon) " and MrSnuggles99 off-line etc.).By clicking menu item (not shown), partner's video clipping is from showing that video switch that recorded or that upload is engraved in the live video of ongoing content in its game when described to the current partner who is just playing the game in host services 210.Therefore, it becomes the game view finder hiving off for partner.If select game and this game of user click of partner, this game will be amplified to full screen, and the user's game that can watch full screen on-the-spot broadcasting and play.

Again, watch the user of partner's game not have the duplicate of game, also do not have the local computing/game console resource for playing this game.It is significant instant that game is watched.

As above previously described, in the time that user plays the game in host services 210, user can " return " and plays and find the video segment that it wishes preservation, and then this video segment is saved in to its User Page.This is called as " self-admiration montage (Brag Clip) ".Video segment 2003 is all the self-admiration montage 2003 that the previous game played from it by KILLHAZARD is preserved.Numeral 2004 shows the viewed how many times of self-admiration montage, and when viewed self-admiration montage is, and user has the chance to its rating, and how high the number of the diagram 2005 of orange keyhole shape instruction grade be.In the time that user watches User Page, self-admiration montage 2003 circulates consistently together with all the other videos on the page.If user selects and click the one in self-admiration montage 2003, it amplifies to present self-admiration montage 2003, and allow to play, suspend, return, the DVR of this montage such as F.F., stepping controls.

Self-admiration montage 2003 playback are be stored in the compressed video segment on RAID array 1511-1512 and its decompress(ion) is contractd its playback is implemented while being written into the montage of user record self-admiration by application program/game server 1521-1525.

Self-admiration montage 2003 also can be " 3D DVR " video segment (also, from the game state sequence that can be reproduced and allow user to change camera to watch game a little) of the game of self-supporting 3D DVR ability.Under this situation, the compressed videograph of specific " the leaping " of carrying out when the recording game fragment except user, also stores games state information.When all circulation times consistently of the just viewed and all thumbnail of User Page and video window, 3D DVR self-admiration montage 2003 will circulate the self-admiration montage 2003 that is recorded as compressed video when " leaping " of user record game fragment consistently.But in the time that user selects 3D DVR self-admiration montage 2003 and clicks the montage 2003 of 3D DVR self-admiration, except allowing to play the DVR control of compressed video self-admiration montage, user can click and provide its button for the 3D DVR ability of the fragment of playing.Its camera that can control independently during game fragment " leaps ", if and it wishes (and the user who has User Page allows so), it can record substituting self-admiration montage with the form of compressed video and " leap ", substituting self-admiration montage " leapt " (immediately, or after the owner of User Page has the chance of verifying self-admiration montage) used other beholders that then can be User Page.

These 3D DVR self-admiration montage 2003 abilities are that the game by start the game state information that recorded will reset on Another Application program/game server 1521-1525 is enabled.Because game can almost be started (as described previously) instantaneously, thus start its (its broadcasting is limited to the game state by self-admiration compilation film segment record) and then allow user by compressed videograph in delay buffer 1515 with camera carry out " leaping " not difficult.Carry out " leaping " once user completes, game is cancelled to startup.

From user's viewpoint, start " the leaping " with 3D DVR self-admiration montage 2003 unlike the DVR control difficulty of controlling linear self-admiration montage 2003.User can not know this game or does not even know how to play this game.User indicates the virtual camera operator in the 3D world during the game fragment of staring at another operator's record.

User records in self-admiration montage (or record or upload from microphone) by also the audio frequency of himself being added.By this way, can use from personage and the action of game and produce custom tailored motion picture with self-admiration montage.This animation design technique is commonly called " game film (machinima) ".

Along with user makes progress in game, it will reach different technical ability ranks.The game of playing is by achievement Report to operation control system 401, and described technical ability rank also will be shown on User Page.

Interactive animated advertising

Online advertisement is converted to still image, video from word, and is converted to now interactive fragment, conventionally with implementing as the animation thin client of Adobe Flash.Use the reason of animation thin client to be: user is conventionally for being delayed because of the privilege of promoting the sale of products to it or serve without patient.And thin client is carried out on the PC of unusual low performance, and therefore advertiser can have height confidence: interactive type advertisement will suitably be worked.Regret, upper restricted in the duration of interactive degree and experience (to reduce download time) such as the animation thin client of Adobe Flash.

Figure 21 illustrates an interactive type advertisement, and wherein user will select the outside and inner color of automobile in the time that automobile rotates in showroom, and how ray tracing display automobile seems in real time simultaneously.Then user selects role to drive a car, and then user can adopt this automobile to come for driving on contest track or through the foreign place such as Monaco.User can select larger engine or better tire, and then can see how the configuration of change affects automobile acceleration or keep stable ability.

Certainly, advertisement effectively be most advanced and sophisticated 3D video-game.But for this advertisement that can play on PC or video game console, it need to be downloaded by 100MB, and under the situation of PC, it may need to install special driver, and may in the time that PC lacks enough CPU or GPU computing capability, not carry out.Therefore, described advertisement is unrealistic in previous technical configuration.

In host services 210, described advertisement is almost thrown in immediately, and preferably carries out, regardless of user's client 415 abilities.Therefore, it is more promptly thrown in than thin client interactive type advertisement, abundanter in experience, and highly reliable.

Mobile geometry during real-time animation

RAID array 1511-1512 and enter port route 1502 and so fast data rate can be provided and there is so low time delay, so that likely design dependency is in RAID array 1511-1512 and for example enter port route 1502, in the video-game and the application program that directly transmit reliably geometry during real-time animation (, having leaping of complex database) in the centre of game play or application program.

The system of prior art (such as, shown video game system in Fig. 1) under, available mass storage equipment (especially in practical household equipment) is extremely slowly so that can not be in flow geometry during game play (except the predictable a little situation of required geometry).For example, in the driving game that has particular link, can reasonably suitably predict for geometry and the mass storage equipment of the building in coming into view and can search in advance bit position that geometry on the horizon is determined.

But in the complex scene with uncertain change (for example, have around in complicated personage's campaign scene), if the RAM on PC or video game system is filled the geometry for the current object at view completely, and then what user is after suddenly its personage being turned to watch its personage, if geometry is not written in RAM in advance, may before geometry can be shown, there is delay.

In host services 210, the speed that RAID array 1511-1512 can exceed ultrahigh speed ethernet speed flows data, and under SAN network, likely reach the speed of 10,000,000,000 bps that is better than 10 1,000,000,000 bit Ethernets or is better than other network technologies.10000000000 bps by being less than the data that are written into gigabyte in a second.In 60fps frame time (16.67 milliseconds), can be written into the data of approximately 170 megabits (21MB).Certainly, even, in RAID configuration, rotation media also will cause being greater than the time delay of a frame time, but taking flash memory as basic RAID holder general large with rotation media RAID array and will can not cause this high time delay the most at last.In each embodiment, use the high-speed cache writing via a large amount of RAM that the very access of low delay is provided.

Therefore, at sufficiently high network speed, and under the mass storage of enough low delays, can process with CPU and/or GPU that 3D data are general fast to be flow to geometry in application program/game server 1521-1525.Therefore, in previously given example, wherein user turns to its personage and see backward suddenly, can before completing rotation, personage be written into its all personages' after one's death geometry, and therefore, for user, by appear its in the world of the general real photo realistic of live action in.

As discussed previously, the one in the last border in the computer animation of photo realistic is face, and because human eye is for the sensitiveness of imperfection, can cause the negative reaction from beholder from the facial small fault of photo realistic.Figure 22 shows use Contour ^tMauthenticity capture technique (the theme of the application in following common co-pending application: the 10/942nd of application on September 15th, 2004, No. 609 " Apparatus and method for capturing the motion of a performer "; The 10/942nd of application on September 15th, 2004, No. 413 " Apparatus and method for capturing theexpression of a performer "; The 11/066th of application on February 25th, 2005, No. 954 " Apparatusand method for improving marker identification within a motion capturesystem "; The 11/077th of application on March 10th, 2005, No. 628 " Apparatus and method forperforming motion capture using shutter synchronization "; The 11/255th of application on October 20th, 2005, No. 854 " Apparatus and method for performing motion captureusing a random pattern on capture surfaces "; The 11/449th of application on June 7th, 2006, No. 131 " System and method for performing motion capture usingphosphor application techniques "; The 11/449th of application on June 7th, 2006, No. 043 " System and method for performing motion capture by strobing a fluorescentlamp "; The 11/449th of application on June 7th, 2006, No. 127 " System and method for threedimensional capture of stop-motion animated characters ", each in above-mentioned application is all transferred to the assignee of this CIP application) how the live performance of catching to cause the very level and smooth surface of catching, afterwards reach the tracking surface (also, facial motion is accurately followed in polygon motion) of high polygon counting.Finally, in the time the video of live performance being mapped in to tracking surface above with generation grain surface, produce the result of photo realistic.

Although current GPU technology can be reproduced many polygons and this surface of throwing light in real time of following the trail of in surface and texture, if but each frame time of polygon and texture changes (it will produce the result of tool photo realistic sense), it will promptly consume all available RAM of modern PC or video game console.

Use flow geometry technology as described above, thus by geometry be constantly fed in application program/game server 1521-1525 with make its constantly the face of cartoon making photo realistic allow to produce to have and almost can not be different from the facial facial video-game of live action and become actual.

The integration of linear content and interactive feature

Film, TV programme and audio material (being referred to as " linear content ") can be used for family and office user with many forms widely.Linear content can be obtained on as the tangible media of CD, DVD, HD-DVD and blue light media.It also can be by carrying out record from the DVR of satellite and cable tv broadcast.In addition, it can be via satellite and the pay-per-view of cable tv (PPV) content and available with the video request program on cable tv (VOD).

The linear content day by day increasing can be used with content and the mobile content downloaded via the Internet.Now, really do not exist one can experience the characteristic position relevant to linear media.For example, DVD and other video lights media have conventionally in other positions disabled interactive feature (as comment, " titbit " short-movie etc. of director).Online Music website has disabled Album Cover Art and song information on CD conventionally, but is not that all CD are available online.And the website being associated with TV programme usually has additional features, blog and sometimes from performer or creator's comment.

In addition, the in the situation that of many films or motion event, conventionally often there is the video-game of issuing (under the situation of film) or (under the situation of motion) and can closely be related to real world event (for example, player's transaction) together with linear media.

Host services 210 is very suitable for transmitting linear content in the time that the related content of disparate form is linked together.Really, transmitting film not as transmitting the video-game of high interaction formula has challenge, and host services 210 can be sent to the plurality of devices in family or office by linear content, or is sent to mobile device.Figure 23 shows the exemplary user interface page for host services 210, and it shows the selection of linear content.

But, be different from most of linear content distribution systems, host services 210 can also transmit relevant interactive composition (for example, the interactive superimposed layer on the menu on DVD and feature, HD-DVD, and Adobe Flash animation (as mentioned below) on website).Therefore, client device 415 limits and no longer introduces the restriction which feature can be used.

In addition, Co-location system 210 can dynamically and in real time link together linear content and video game content.For example, if user is just watching Quidditch (Quidditch) match in Harry Potter's film, and determine that it is ready to attempt playing Quidditch, it can only be clicked button and film and will suspend and it will be transported to the Quidditch fragment of Harry Potter's video-game immediately.After object for appreciation Quidditch match, button is clicked in another time, and film will restart at once.

In the case of the figure and manufacturing technology of photo realistic, the video that wherein photography is caught can not be different from live action figure, in the time that user carries out the transformation of the Quidditch game in video-game that the Quidditch in live action movie plays to host services (as described herein), in fact these two scenes can not be distinguished.This is that linear content and both directors of interactive (for example, video-game) content provide brand-new creation option, can not distinguish because the line between this two worlds becomes.

Utilize host services framework shown in Figure 14, the control of the virtual camera in 3D film can be offered to beholder.For example, in the scene betiding in train, will likely allow beholder to control virtual camera and look about train in the time that story is made progress.The all 3D objects (" assets ") of this supposition in train and the enough computing capability levels of described scene and original film can reproduced are in real time available.

And, there is the available interactive feature stimulating very much in the amusement even producing for non-computer.For example, 2005 films " Pride and Prejudice " have the many scenes in the old Britain mansion being gaily decorated.For specific mansion scene, user can suspend video and then control camera to make an inspection tour mansion, or possible peripheral region.In order to implement these, portability has fish-eye camera through mansion, in the time that it follows the trail of its position, is very similar to the QuickTimeVR that implements prior art Apple (apple) company.Various frames then will be converted, and therefore image is undistorted, and then it is stored in RAID array 1511-1512 above together with film, and be played in the time that user selects to continue virtual tour.

The in the situation that of motion event, can live motion event (such as, Basketball Match) be flowed via host services 210 and watch (want for common TV as it) for user.After user watches specific broadcasting, the video-game (final basketball player seems the general photo realistic with true player) of game can be caught up with the player who starts in same position, and user's (may control separately a player) can play to watch its whether comparable described player to be get Geng Jia again.

Host services 210 described herein is extremely suitable for supporting this future world, because its can bear unrealistic so that can not be installed in family or most of offices background in computing capability and large capacity storage resources, and always up-to-date (in the case of available up-to-date computing hardware) of its computational resource, but in family background, will always exist and there is the PC in older generation and the family of video-game.In addition, in host services 210, user is concealed all these computation complexities, therefore, even if user may just use very most advanced and sophisticated system, sees from user's viewpoint, also simple as changed the channel on TV.In addition, user can all computing capabilitys of access and computing capability by the experience bringing from any client 415.

Multiplayer

Degree for game for multiplayer, game can not only be communicated to application program/game server 1521-1525 via entering port route 1502 networks, and will be conveyed to and be had the not server of execution or the Internet (not shown) of game machine in host services 210 by network bridge.In the time that the computer by common interconnection network is played multiplayer, application program/game server 1521-1525 will have the benefit (compared with situation about carrying out on game server in the family) of the access the Internet that is exceedingly fast, but its will be subject in the capabilities limits that slowly connects other computers of playing games, and be also subject to potentially game server on the Internet to be designed to adapt to the fact restriction of minimum common ground (common denominator) (described game server is the home computer that relatively slow consumer the Internet connects).

But in the time playing multiplayer completely in host services 210 server-centrics, can reach very big difference.Each application program/game server the 1521-1525 that game for user is carried out to Co-location is by with other application program/game server 1521-1525 and with flank speed, utmost point low delay connectivity and huge, very fast storage array, the center-control for multiplayer is carried out any server interconnect of Co-location.For example, if ultrahigh speed Ethernet is used for into port route 1502 networks, application program/game server 1521-1525 will pass in the middle of each other, and be communicated to any server that the center-control for multiplayer is carried out Co-location with 1,000,000,000 bps of speed under the potential only time delay below 1 millisecond or 1 millisecond.In addition, RAID array 1511-1512 can respond very rapidly and follow with 1,000,000,000 bps of data speed transmissions.As an example, if user customizes personage aspect appearance and clothes, to make personage there is a large amount of geometries and behavior unique for personage, be limited under the prior art system of the game client of carrying out in PC or game console in the family, if described personage will enter in another user's the visual field, user will must wait for until long slow download completes, to all geometries and behavioral data are written in its computer.In host services 210, described identical download can be better than the ultrahigh speed Ethernet from RAID array 1511-1512 service with 1,000,000,000 bps of speed.Even if having 8Mbps the Internet, domestic consumer connects (it is exceedingly fast according to standard now), also fast 100 times of ultrahigh speed Ethernets.Therefore, on fast the Internet connects, spend being operated on ultrahigh speed Ethernet of carrying out for a minute and cost is less than to one second.

Top player hives off and championship

Host services 210 is extremely suitable for championship.Because carry out in local client without game, so the chance that does not exist user to practise fraud.And, due to the ability of output route 1540 multicast UDP streams, make host services 210 while to broadcast larger championship to the thousands in spectators.

In fact, when exist be so in fashion so that when thousands of users are just receiving the particular video stream of same stream (for example, show the view of larger championship), can be more effective by video flowing send to content distribution network (CDN) (such as, Akamai (A Kamai company) or Limelight (spotlight company)) for being assigned in a large number many client devices 415.

When show the game view finder page that top player hives off with CDN, can obtain the efficiency of similar level.

For larger championship, can during particular game, provide comment with live famous person announcer.Although a large number of users will be to watch larger championship, and relatively small number order will be to play in championship.Can be by the application program/game server 1521-1525 that is routed to user to playing from famous person announcer's audio frequency carries out Co-location and any onlooker's pattern duplicate of the game in championship is carried out to Co-location in championship, and audio frequency can be added and record on game audio frequency.Can be in game (also may just on onlooker's view) stack famous person announcer's video.

The acceleration that webpage is written into

The main transportation protocol of world wide web, HTML (Hypertext Markup Language) (HTTP) are conceived and are limited at wherein only business and had high speed internet connection, and online consumer used in the epoch of dial data machine or ISDN.Now, be T1 line for " golden standard " that connect fast, it provides 1.5Mbps data rate (also, having identical data speed in both direction) symmetrically.

Now, situation is completely different.The average family connection speed connecting via DSL or cable modem in a large amount of developed worlds has the downstream data rate more much higher than T1 line.In fact,, in some places in the world, Fiber To The Curb, and (fiber-to-the-curb) just brings the data rate up to 50Mbps to 100Mbps into family.

Unfortunately, HTTP is not become to effectively utilize this speed improvement sharply by framework (not also being implemented).Website is the set of the archives on remote server.Very briefly, HTTP asks the first archives, these archives that wait for downloads, and then ask the second archives, these archives etc. wait for downloads.In fact, HTTP allows one above " open connection " (also, asking more than one archives) at every turn, but because the standard of deciding through consultation (and preventing the hope that the webserver is overloaded) makes only to permit considerably less open connection.In addition,, due to the mode that webpage is constructed, browser is not usually recognized the multiple while pages (yet, only just becoming apparent after dissecting a page: need to download as the new archive of image) that can be used for downloading immediately.Therefore, the archives on website are to be in fact one by one written into.In addition,, due to the request being used by HTTP and response protocol, there is the time delay of roughly (typical web servers of Visit the USA) relevant to loaded each archives 100 milliseconds.

In the situation that low speed connects relatively, this can not introduce a large amount of problems, because determine the stand-by period of webpage for the download time of archives itself.But, along with connection speed increases (in especially complicated webpage situation), start to cause problem.

In Figure 24, in shown example, show typical commercial website (this specific website is from larger sport footwear trade mark).On website, there are 54 archives.Archives comprise HTML, CSS, JPEG, PHP, JavaScript and Flash archives, and comprise video content.Live webpage (also, user can click this webpage and bring into use this webpage) before, must be written into 1.5M byte altogether at the scene.There are many reasons for a large amount of archives.First, this webpage is complicated and most advanced and sophisticated webpage, and secondly, this webpage be the webpage that dynamically combines of the information of the user based on about this page of access (for example, user is from which country, whether which kind of language, buy before user etc.), and download different archives depending on all of these factors taken together.But it is still very typical business webpage.

Figure 24 shows along with connection speed increases live webpage elapsed time amount before at the scene.1.5Mbps connection speed 2401 times, use the legacy network server with conventional web browser, before live webpage, spend 13.5 seconds at the scene.12Mbps connection speed 2402 times, be written into time decreased to 6.5 second, or approximately fast again.But 96Mbps connection speed 2403 times, the time that is written into is only reduced to approximately 5.5 seconds.This reason is because under this high speed of download, downloads the minimal time of archives itself, but each archives separately roughly the time delay of 100 milliseconds still keep, thereby cause 54 archives * time delay of 100 milliseconds=5.4 seconds.Therefore, how soon no matter arrive the connection of family, before this website is live at the scene, will always spend at least 5.4 seconds.Another factor is that server side is queued up; Each HTTP request is to add at the rear portion of queue, and therefore on busy server, this will have appreciable impact, because for each small project that will obtain from the webserver, HTTP request need to wait for that it returns.

The mode addressing these problems is discarded or again limits HTTP.Or, may make website owner preferably its archives are merged into single archives (for example,, with Adobe Flash form).But, as a practical problem, the said firm and many other people in its web site architecture, there is great amount of investment.In addition, connect although some families have 12-100Mbps, most of families still have speed more slowly, and HTTP is slowly working well under speed really.

An alternative method is, on application program/game server 1521-1525, web browser is carried out to Co-location, and (or potentially in the RAM on the application program/game server 1521-1525 that web browser is carried out to Co-location or on local holder) carries out Co-location to the archives for the webserver on RAID array 1511-1512.Due to the very fast interconnection via entering port route 1502 (or to local holder), not using every archives under HTTP to there is the time delay of 100 milliseconds, but under use HTTP, will have the minimum time delay of every archives.Then, not make user in family via HTTP accessing web page, but user can be via client 415 accessing web pages.Then, even under 1.5Mbps connects (because webpage for this reason does not need massive band width to come for its video), webpage also will be less than at each line 2400 in time of 1 second in broadcasting live.Before the live page of the web browser displaying scene carried out, will not there is not time delay on application program/game server 1521-1525, and will not have the time delay that can detect before client 415 shows the video output from web browser in fact.In the time that user uses mouse search webpage and/or key in word on webpage, user's input message is sent to the web browser of carrying out on application program/game server 1521-1525, and web browser will correspondingly respond.

A disadvantage of the method is: if compressor reducer transmitting video data just consistently, utilized bandwidth, also like this even if webpage becomes static state.This can by configuration compressor reducer with only (and if) just transmit data and then only transfer data to the page part changing when web-page change and make up.In the time that existence has consistently some webpages of the flicker poster that changes etc., described webpage is often horrible, unless and have the reason of wanting move something (for example, video clipping), otherwise webpage is static conventionally.For described webpage, may be following situation: use host services 210 will transmit less data (compared with legacy network server), because will only transmit the image of actual displayed, there is no thin client executable code, and large object that cannot be never viewed (such as, rolling flip image).

Therefore, by host services 210, old edition webpage is carried out to Co-location, webpage can be written into time decreased to turning on the degree that webpage is the channel on similar change TV: this webpage of effectively broadcasting live immediately.

Be convenient to the debug of game and application program

As discussed previously, there is the video-game of real-time graph and application program and be very complicated application program and conventionally in the time that it is issued in this field, it is containing defectiveness.Although software developer will obtain the feedback about defect from user; and it may have some modes of machine state being passed back for after collapse, but identification is what causes game or real-time application collapse or carries out inadequately very difficult definitely.

In the time playing or application program carries out in host services 210, the video/audio output of game or application program is recorded on delay buffer 1515 consistently.In addition, house dog process is carried out each application program/game server 1521-1525, this house dog process will to host services control system 401 termly reporting application/game server 1521-1525 carry out just reposefully.If house dog process is failed report, server control system 401 will be attempted to communicate by letter with application program/game server 1521-1525, and if success, by the available whatsoever machine state of collection.By whatsoever can with information send to software developer together with the video/audio being recorded by delay buffer 1515.

Therefore,, when game or application software developer are in the time that host services 210 obtains the notice of collapse, it obtains causing the frame of the reason of collapse to connect the record of frame.This information is tracking defect and will may be valuable in defect repair.

Should also be noted that in the time that application program/game server 1521-1525 collapses, restart server in the nearest moment of restarting, and message is offered to user, thereby apologize with regard to technical difficulty.

Resource sharing and cost savings

In Fig. 4 a and Fig. 4 b, shown system provides multiple benefit for terminal use and game and application developers.For example, conventionally, family and office's FTP client FTP (for example, PC or game console) only the little percentage in one week hour in be in use.According to by Nielsen (Nelson) amusement " Active Gamer Benchmark Study (enliven player's datum mark research) " (http://www.prnewswire.com/cgi-bin/stories.pl? ACCT=104 & STORY=/www/story/10-05-2006/0004446115 & EDATE=) 5 days October in 2006 communication original text, active player's average cost in a week 14 hours object for appreciation and playing in handheld device for approximately a week 17 hours on video game console.This report is also stated: for all game play activities (comprising control desk, handheld device and PC game play), average one week 13 hours of active player.Consider the control desk video-game reproduction time of higher number, have one week 24*7=168 hour, this hint enliven player in, video game console only the 17/168=10% of a week hour in be in use.Or 90% time, video game console leaves unused.Given video game console expensive, and manufacturer subsidizes the fact of described equipment, this is the very inefficient use of expensive resource.PC in business also only uses conventionally in the part of a week hour, especially usually required non-portable Desktop PC of high-end applications program (such as, Autodesk Maya).Although some business were at all hours and all operate holiday, and some PC (for example, take the portable PC for carrying out at night work home) be all hours and use holiday, but most of business activities tend to concentrate in given business time zone from the approximately 9AM to 5PM of Mon-Fri, less holiday and time of having a rest (such as, lunch), and because using, most of PC occurs in the time that user utilizes PC energetically, so it is followed: the utilization of Desktop PC tends to follow these hours run numbers.If suppose in one week five days constantly use PC from 9AM to 5PM, this by hint PC the 40/168=24% of a week hour in used.High-performance Desktop PC is the very expensive investment for business, and this has reflected low-down availability.The school imparting knowledge to students on desktop computer can be used computer in the even more fraction of a week, although and its hourage depending on teaching and changing, great majority teaching from Mon-Fri in the daytime hour during occur.Therefore, generally speaking, PC and video game console are only utilized in the fraction of a week hour.

Notably, because many people the Mon-Fri of eekday in the daytime hour during in business or at Schools Work, so these people are playing video game not during these hours conventionally, and therefore in the time of its certain playing video game, its normally during other hours (such as, evening, weekend and holiday).

The configuration of shown host services in Given Graph 4a, the use pattern described in above-mentioned two sections causes the very effective utilization of resource.Obviously, there is the restriction for the user's that can be served by host services 210 in preset time number, the real-time responsiveness that especially need to be used for complex application (as most advanced and sophisticated 3D video-game) in the case of user.But, the PC (it is conventionally in the idle placement of most of time) that is different from the video game console in family or is used by business, server 402 can be re-used at different time by different user.For example, the high-performance server 402 with the two CPU of high-performance and two GPU and a large amount of RAM can be utilized at the 9AM to 5PM of eekday by business and school, but by the player who plays most advanced and sophisticated video-game at night, weekend and holiday utilize.Similarly, low performance application program can be utilized on Celeron (Celeron) CPU, low performance server 402 without GPU (or very the GPU of low side) and limited RAM and low performance game can utilize low performance server 402 during non-commercial hour having during business hour by business and school.

In addition, in the situation of described host services configuration, resource is effectively to share in the middle of thousands of names (if not millions of name) user in this article.Generally speaking the little percentage that, online service only has its total user base is in use service preset time.If consider previously listed Nielsen video-game usage statistics, what be easily appreciated that.If enliven player's only object for appreciation control desk game in 17 hours in a week, if and the peak usage time of make is (5-12AM at night, 7*5 days=35 hours/week) and weekend (8AM-12AM, 16*2=32 hour/week) typical inoperative, non-commercial hour during,, for the game play of 17 hours, within one week, there is 35+32=65 peak value hour.Due to following many former thereby make to be difficult to the definite peak user load on estimating system: some users will time durations object for appreciation peak value outside, may exist time between particular day to have user's (clustering) peak value of gathering together, time to peak can be subject to the impacts such as played games type (for example, child's game may be early time object for appreciation at night).But, suppose in the time that player may play games, player plays mean hours number much smaller than hourage in the daytime, only the user of a part of number of host services 210 uses host services 210 in preset time.For this analysis, we suppose that peak load is 12.5%.Therefore, only 12.5% calculating, compression and bandwidth resources are to use in preset time, thereby support given user to play the given rank of performance game because re-using of resource causes only 12.5% hardware cost.

In addition, suppose the computing capability that some game and application program need to be more than its other, game that can be based on being played by user or the application program of being carried out by user are carried out dynamically Resources allocation.Therefore, select the user of low performance game or application program will be assigned with low performance (cheaper) server 402, and select the user of high-performance game or application program will be assigned with high-performance (more expensive) server 402.In fact, given game or application program may have lower-performance and the superior performance district of game or application program, and can between the district of game or application program, user be switched to another server 402 from a server 402, to keep user to carry out on the least cost server 402 of needs that meets game or application program.Note, can be by even low performance server 402 is used than the RAID array 405 that single disk is faster, this has the benefit of very fast disk transfer rate.Therefore, every server 402 average unit costs of crossing over all application programs of playing games or use are more much smaller than the cost of the most of expensive server 402 of the game of object for appreciation peak performance or application program, but, even if low performance server 402 also can obtain disk performance benefit from RAID array 405.

In addition, the server 402 in host services 210 may be the PC motherboard without disk or perimeter interface (being different from network interface), and just, can be integrated into the one single chip just having to the fast network interface of SAN 403 downwards.And RAID array 405 may will share in the middle of the user much more situation than there is disk, therefore each active user's disk cost will be much smaller than a disc driver.All this equipment may reside on environment in the support in check room server environment.If server 402 is out of order, it can easily place under repair or replace at host services 210 places.By contrast, PC in family or office or game console must be firm, must be able to survive rational wearing and tearing and tear in case thumped or the independent utensil that lands needs shell, there is at least one disc driver, must survive hostile environment condition (for example, being filled in reluctantly in the overheated AV cabinet with other apparatus), need service guarantees, must be packed and shipment, and sold by the retailer that may collect retail interest.In addition, PC or game console must be configured to meet expection game the most intensive in the calculating using in a certain moment in future or the peak performance of application program, even if lower-performance game or application program (or district of game or application program) also may be played at most of time.In addition,, if PC or control desk are out of order, making it obtain repairing is expensive and process (adversely affecting manufacturer, user and software developer) consuming time.

Therefore, suppose that the experience of the experience that is equivalent to local computing resource is offered user by system shown in Fig. 4 a, in family, office or school, experience the computing capability of given level for user, provide described computing capability to want much cheap by framework shown in Fig. 4 a.

Eliminate the needs to upgrading

In addition, user needn't worry the new application program that PC and/or control desk are upgraded to play new game or processed superior performance again.It is used that any game in host services 210 or application program (no matter described game or application program need the server 402 of what type) all can be user, and all game and application program approach to carry out immediately, local holder from RAID array 405 or server 402 is written into rapidly) and suitably there is latest update and defect repair (is also, software developer can select the desired server configuration of the server 402 for carrying out given game or application program, and then server 402 is disposed to optimal drive device, and then As time goes on, developer can will upgrade simultaneously, defect repair etc. offer game in host services 210 or all duplicates of application program).In fact, after user brings into use host services 210, user may find that game and application program (for example continue to provide better experience, via upgrading and/or defect repair) and may be following situation: user finds after 1 year that new game or application program (for example can be used for utilizing computing technique, the GPU of superior performance) in the service 210 of (its year before not even exist), therefore for user, can not buy and will after 1 year, play games or the technology the year before of executive utility.Because play games or the computational resource of executive utility is invisible for user (is also, viewpoint from user is seen, user only selects to start to approach the game carried out immediately or application program-more change the channel on TV as user), thus user's hardware by the situation that user does not even recognize upgrading by " upgrading ".

Eliminate the needs for backup

Backup for another the larger problem of the user in business, school and family.If disk is out of order, if or exist and be not intended to wipe, being stored in information in local PC or video game console (for example, under the situation of control desk, user's game achievement and grade) may lose.Existence is provided for the many available application program of backup manually or automatically of PC, and game console state can be uploaded to line server for backup, but conventionally local backup is copied to another local disk (or other non-volatile storage facilitiess) that must be stored in safe and organized somewhere, and makes for the backup of online service usually limited owing to connecting available slow speed uplink via typical low-cost Internet.The host services of Fig. 4 a 210 times, the data that are stored in RAID array 405 can configure with the prior art RAID configuring technical being well known to those skilled in the art, so that proper disk is while being out of order, not obliterated data, and notice is being held to the technician at server-centric place of out of order disk, and then technician will replace this disk, and this disk then will automatically be upgraded to make RAID array to tolerate again fault.In addition, because all disc drivers are closer to each other and have the quick local network via SAN 403 therebetween, so it is not difficult in server-centric, all disk system configurations to be backuped to secondary holder (it can be stored in server-centric place or reorientate through changing places) termly.From the user's of host services 210 viewpoint, its data overall safety all the time, and it is from considering backup.

To the access of demonstration

User often wishes in front trial game or the application program of buying game or application program.As discussed previously, there is prior art apparatus, demonstrate by this prior art apparatus that (verb form of " demonstration " meaning is crippled version on probation, crippled version is also referred to as " demonstration ", but as noun) game and application program, but wherein each is limited and/or is not convenient.Use host services 210, for user, easily and be convenient to demonstration on probation.In fact, what user carried out is selects demonstration and this demonstration on probation via user interface (such as, below described user interface).Demonstration will almost be written on the server 402 that is suitable for this demonstration immediately, and it carries out any other game of all fours or application program.No matter demonstration needs very high performance server 402 or the server of low performance 402, and what type family or office's client 415 that no matter user uses are, sees from user's viewpoint, demonstrates all by work.The software publisher of game demonstration or application demo can be controlled definitely and permit user's what demonstration on probation and how long on probation, and certainly, demonstration can be included as user provides the user interface key element of the access chance that obtains the full version to demonstrated game or application program.

Because demonstration may be lower than cost price or provide free, some users may attempt to use the demonstration (especially repeatedly playing the game demonstration that possibility is interesting) repeating.Host services 210 can limit for given user's demonstration and use by various technology.The most direct method is to set up for each user's user ID and restriction to allow given user ID to play the number of times of demonstration.But user can arrange multiple user ID, especially it is freely in situation.That restriction allows given client 415 to play the number of times of demonstration for a technology of head it off.If client is autonomous device, this equipment will have a sequence number, and host services 210 can limit demonstration can be by the number of times of client access with described sequence number.If client 415 is just carried out with the software on PC or other equipment, can be assigned sequence number and this sequence number is stored in to PC by host services 210 goes up and uses this sequence number to limit demonstration use, but supposition PC can carry out sequencing again by user, and sequence number is wiped free of or changes, another option is that host services 210 keeps the record of PC network adapter media interviews controls (MAC) address (and/or other machines specific identifiers, such as hard disk drive sequence number etc.) and demonstration is used and is limited to this MAC Address.Suppose the MAC Address that can change network adapter, but this is not open-and-shut method.Other method is the number of times that restriction demonstration can be played to given IP address.Although can come periodically again to assign IP address by cable modem and DSL supplier, but it can not occur in practice very continually, for example, and if (can determine, by contact ISP) IP is the block in the IP address for house DSL or cable modem access, conventionally can set up for the demonstration of the peanut of given family and use.And, after the NAT router that shares same IP address, may there is multiple equipment in the family, but conventionally in house background, will have equipment described in limited number.If IP address is in the block in service business, can set up the demonstration for the greater number of business.But last, the combination of all previous described methods is best modes of the number of the demonstration on restriction PC.Although may not there is not the open-and-shut mode that determined and technical skilled user may be restricted in the number of repeat playing demonstration, but setting up a large amount of obstacles can set up enough obstructions and abuse demo system to make most of pc users be unworthy taking great energy, and contrary, it uses demonstration in the time that it is intended to new game on probation and application program.

To the benefit of school, business and other mechanisms

Significantly benefit especially appears at the business, school and other mechanisms that utilize system shown in Fig. 4 a.The essence cost that business and school have and installation, maintenance and upgrading PC are associated, especially in the time referring to the PC of high performance application of execution such as Maya.As stated previously, PC conventionally only one week hour a part in be utilized, and as in the family, the cost of the PC of performance capability with given level in office or school environment is far above the cost in server-centric environment.

For example, compared with the situation of large-scale commerce or school's (, large university) under, the IT department of described entity arrange server-centric and safeguard via LAN level connect and remotely the computer of access can be actual.Exist for connecting and many solutions of remote access computer via LAN or via the private high bandwidth between office.For example, by the Windows Terminal server of Microsoft, or by virtual network computing application program (as the VNC from RealVNC (Long-distance Control) Co., Ltd) or by the thin client device from Sun Microsystems (sun computer system house), user can obtain the remote access to PC or server, has the quality of certain limit in graphic response time and user's experience.In addition, the described server-centric of management is voluntarily exclusively used in single business or school conventionally, and for example therefore can not utilize, disparate application program (, amusement and business applications) possible use overlapping in the time that the different time of a week utilizes same computational resource.Therefore, much business and school lack independent scale, resource or the technical skill with the server-centric connecting to the network of each user's LAN speed of arranging.In fact the school of percentage and business have the Internet connection (for example, DSL, cable modem) identical with family, greatly.

But, the described needs that still may have for very high performance calculating (or termly or periodically) of organizing.For example, Minor Construction company may only have the architect of peanut, in the time carrying out design work, has the calculating needs of appropriateness relatively, but it may periodically need very high performance 3D to calculate (for example,, in the time that foundation leaps for the 3D of the new building design of client).System shown in Fig. 4 a is extremely suitable for described tissue.Described organizing only needs to connect (for example, DSL, cable modem) and conventionally very cheap for providing to the network of the one species of family.It can utilize cheap PC as client 415, or does not have PC also passable completely, and utilizes the simple cheap special equipment of implementing control signal logic 413 and low delay video decompression 412.This feature is attractive especially for the school of problem may with the stealing of PC or the damage to the personal module in PC.

This configuration has solved the many problems (and many this advantages are also for the domestic consumer who carries out general-purpose computations shares) for described tissue.For example, running cost (it finally must be back to user to have feasible business with certain form transmission) may be much lower, because (a) computational resource is to share from other application programs in a week with different peak usage times, (b) described tissue can only obtain (and causing cost) access to high-performance calculation resource when needed, and (c) described tissue needn't be provided for backing up or otherwise safeguarding the resource of high-performance calculation resource.

Pirate elimination

In addition, game, application program, interactive film etc. may be for another example now like this by piracies.Because game is to carry out at server-centric place, so user does not possess the access for base program code, therefore there is not piracy.Even if user will replicating original code, user can not carry out this code on standard game control desk or home computer.This has opened the normal video game market of disabled all over the world (such as, China).Again the sale of the game having used is also impossible.

For game developer, as situation now, there is less city field's discontinuity.Force user and developer's upgrading and game developer to depend on the current situation contrast of the timely transmission of hardware platform with a brand-new generation technique, can be along with the time be along with game requires change and little by little upgrades host services 210.

Streaming interactive video

More than describe and provide by (it impliedly also comprises the audio frequency together with video, the multiple application of novel basic conception activation as used herein) taking common interconnection network as basic low-latency streaming interactive video.Provide the prior art system of the video that flows only to there is the application of the institute's activation that can implement by high time delay interaction via the Internet.For example, for example, suitably work under high time delay for the basic playback controls of linear video (, suspend, return, F.F.), and likely in the middle of linear video feed, select.In addition, as stated previously, the character of some video-games allows it to play with high time delay.But, the high time delay (or low compression ratio rate) that is used for the prior art approach of mobile video seriously limits the potential application of the video that flows or makes its deployment narrow to specialized network environment, and even, in described environment, prior art is also introduced the essence burden on network.Technology described herein has been opened the gate of multiple application possible under the low-latency streaming interactive video via the Internet, especially connects and the described application of activation via consumer's level the Internet.

In fact, with the general little client device of the client 465 of Fig. 4 c under, be enough to the network that is exceedingly fast between quick storage and the powerful server of computing capability by effective any amount, any amount and connect and provide the user of enhancing to experience, it makes the new calculating epoch become possibility.In addition, because bandwidth requirement does not increase along with the computing capability of system increases, because bandwidth requirement is only about display resolution, quality and frame rate), once so broadband internet connectivity be ubiquitous (for example, contain via widespread low delay is wireless), reliably and there is the sufficiently high bandwidth of the needs that are enough to the display device 422 that meets all users, problem will be the application of typical consumer and business necessary be thick and heavy client (such as, carry out Windows, Linux, the PC of OSX etc. or mobile phone) still even thin client (such as, Adobe Flash or Java).

The appearance of streaming interactive video causes rethinking about the supposition of the structure of computing architecture.This example is host services 210 server-centric embodiment shown in Figure 15.Be feedback loop for the video path that postpones to cushion G device and/or hive off video 1550, wherein the output of the streaming interactive video through multicast of application program/game server 1521-1525 is fed back in application program/game server 1521-1525 in real time or via path 1551 via path 1552 after selectable delay.This make by prior art server or local computing framework will be impossible or infeasible multiple practical applications (for example,, such as Figure 16, Figure 17 and application illustrated in fig. 20) become possibility.But as architectural features more generally, feedback loop 1550 recurrence under streaming interactive video level that provides, because can ad infinitum circulate video in the time that application program needs video.This make before never can with multiple application possibility become possibility.

Another crucial architectural features is: video flowing is unidirectional UDP stream.This realizes the multicast (by contrast, such as the bidirectional flow of TCP/IP stream, the number increase along with user being produced to increasing communication stagnation on the network from communication back and forth) of any degree of streaming interactive video effectively.Multicast is the significant capability in server-centric, because the needs of the growth of its permission system to Internet user's (and in fact, the population in the world) respond with at one-to-many or even communicate by letter on multi-to-multi basis.Again, the example of the explanation streaming interactive video recurrence discussed herein and the use of multicast (such as, Figure 16) be only the tip on the potential very large iceberg of tool.

In one embodiment, illustrated various functional module group and the step that is associated can be by containing being useful on the specific hardware components (such as, special purpose integrated circuit (" ASIC ")) of the hardwired logic of carrying out described step or being carried out by any combination of the computer module being programmed and custom hardware components herein.

In one embodiment, described module can be implemented on TMS320x framework such as Texas instrument (for example, TMS320C6000, TMS320C5000 ... etc.) programmable digital signal processor (" DSP ") on.Can use various DSP, still observe described general principle simultaneously.

Embodiment can comprise the various steps as set forth above.Described step can be embodied in the machine-executable instruction that causes universal or special processor execution particular step.By with the irrelevant various elements of these general principles (such as, computer storage, Hard Disk Drive, input equipment) from figure, save the related fields of avoiding confusion.

The key element of disclosed subject matter also can be used as for the machine-readable medium of store machine executable instruction and provides.Machine-readable medium can include, but is not limited to flash memory, laser disc, CD-ROM, DVDROM, RAM, EPROM, EEPROM, magnetic or optical card, communications media or be suitable for the machine-readable medium of the other types of stored electrons instruction.For example, the present invention can be used as computer program and downloads, this computer program can be via communication link (for example, modem or network connect) by means of being embodied in data-signal in carrier wave or other communications medias from remote computer (for example, server) be sent to requesting computer (for example, client).

Will also be understood that, the key element of disclosed subject matter also can be used as computer program and provides, this computer program can be included in the machine-readable medium that stores instruction above, described instruction can be used for programmed computer (for example, processor or other electronic equipments) to carry out a series of operations.Or described operation can be carried out by the combination of hardware and software.Machine-readable medium can include, but is not limited to floppy disk, CD, CD-ROM, and magneto optical disk, ROM, RAM, EPROM, EEPROM, magnetic or optical card, communications media or be suitable for the media/machine-readable medium of the other types of stored electrons instruction.For example, the key element of disclosed subject matter can be used as computer program and downloads, its Program can for example, be sent to request process by means of being embodied in the data-signal in carrier wave or other communications medias from remote computer or electronic equipment via communication link (, modem or network connect).

In addition, although describe disclosed subject matter in conjunction with specific embodiment, numerous modifications and changes will be suitably in category of the present disclosure.Therefore, specification and graphicly should be considered as illustrative and nonrestrictive meaning.

Claims (16)

1. for an equipment for the view in conjunction with multiple real-time streaming interactive videos, this equipment comprises:

Multiple servers, the plurality of server is configured to move one or more very fast video-games or application program to generate multiple unpressed streaming interactive video stream; And

Compression unit, one or more in the unpressed streaming interactive video stream being generated by described server are compressed into new compressed streaming interactive video stream by this compression unit, the described new compressed streaming interactive video stream of described compression unit output is transferred to connect by network with packetizing form the multiple client devices that are associated with corresponding multiple users, at least one user provides control inputs at least one server in described server, the place that described at least one user is geographically positioned at away from least one server in described server,

Wherein with at 1500 miles or be less than in the transmission range of 1500 miles from user's control inputs to the described new compressed streaming interactive video stream of round trip time delay compression that the response of this user's control inputs is presented to user's client device 90 milliseconds of situations worst.

2. equipment according to claim 1, described in wherein one or more, server also operationally produces graphic user interface, and this graphic user interface allows to handle unpressed streaming interactive video stream by user's real-time interactive.

3. equipment according to claim 1, wherein said network comprises the Internet.

4. equipment according to claim 1, wherein said network comprises the Internet, and described at least one, client device comprises the Set Top Box that is operably connected to described the Internet.

5. equipment according to claim 1, wherein said network comprises the Internet, and described at least one, client device comprises the mobile device that is operationally wirelessly connected to described the Internet.

6. equipment according to claim 1, wherein said at least one server allows at least one user to watch the video-game being delayed.

7. equipment according to claim 1, wherein said new compressed streaming interactive video stream comprise for show from one or more described very fast video-games or application program at least one thumbnail video window of mobile video.

8. equipment according to claim 2, wherein said real-time interactive is handled the variation of the view that comprises video-game.

9. equipment according to claim 8, wherein comprises the view of the delay of described video-game in the variation of described view.

10. equipment according to claim 8, wherein comprises the camera view of described video-game in the variation of described view, described camera view is different from least one other user's view.

11. equipment according to claim 2, wherein said graphic user interface provides menu setecting, and this menu setecting is for allowing described user to select the selection of the game to watch simultaneously based on one or more standards.

12. equipment according to claim 11, wherein said one or more standards comprise the type of the video-game of playing.

13. equipment according to claim 11, wherein said one or more standards comprise player's grade.

14. equipment according to claim 11, wherein said one or more standards comprise the total statistics that comprises beholder.

15. equipment according to claim 2, wherein said real-time interactive is handled the change that is included in the audio mix being associated with described one or more video-games or application program.

16. equipment according to claim 14, this equipment also comprises the large capacity holder of being shared by described multiple servers, described large capacity holder is used for storing described statistics.